The OxChief RC is available for 6 Bad Boy mower models:

• MZ Rambler
• MZ Magnum
• ZT Avenger
• ZT Elite
• Maverick
• Maverick HD

I hope that you won’t mind me introducing the component by way of a story.

Boys Will Be Boys

My boys are gone to camp this week. I was excited thinking that the quiet house would yield a utopia of productivity, cleanliness and order. We have been getting things done, and the house is cleaner. But it’s also too quiet around here, and I miss the boys.

Does a man ever outgrow the desire for something RC?

Above, one of the boy’s RC trucks sits atop a Bad Boy Rambler that has been retrofitted with the OxChief RC bolt-on system.

If you’re lucky enough to have a son (or a nephew, or a grandson, or a young guy that you mentor) then perhaps, like me, you often look at their toys with a touch of envy.

• Gel Blasters — where were these when we were kids?
• Go karts — we somehow survived without roll cages and seat belts.
• Dirt bikes — insane fun with a little danger mixed in.
• RC cars — pure fun until the battery runs out (or catastrophic crash).

I still remember looking at RC car magazines when I was a kid, wondering what the cars in pages were really capable of. Were the speed numbers quoted accurate? What did they sound like? Could they pull a wheelie? What did driving them feel like? There’s something about driving a vehicle remotely that tends to put a smile on your face. It’s better than fun, it’s pure fun.

Pure Fun 

There are a handful of things that are simply pure fun — exciting, adventure, and they leave you better off afterwards. For me, this includes snow skiing in Montana, jet skiing anywhere the water is warm, and watching Auburn football. One item in the prior list doesn’t belong — but it’s not because I don’t love Auburn.

But somewhere there’s apparently some memo that at some age, we’re supposed to identify as “Adult”, and this marks the point where we’re supposed to:

• Hold long conversations with others about our fav organic food choices that no-one can pronounce
• Complain ad-nauseam about politics
• Prefer watching screens inside to sweating outdoors
• Wonder aimlessly about lost time rather than imagining a fantastic future

Here’s trusting that you, the reader of this blog, like me, are working to reject the list above.

Quite often, there’s a kernel in the fun that we enjoyed as kids that is far superior to the low-T borefests we settle for as adults.

Don’t settle for binging Netflix when the joy of youth is at hand.

Adding a super solid RC control system to your mower really is a refreshing blast from the fountain of youth.

Why RC my mower?

I think you’ll find that there’s something very satisfying about driving an RC mower — perhaps it’s tied to something we enjoyed long ago that we’ve never really outgrown.

There are several compelling benefits to an RC mower, including:

• Mow from the comfort of your living room regardless of weather
• Tackle dangerous hills / suspect terrain without bodily risk
• Scratch your long-dormant itch to drive an RC vehicle
• Perfect platform to equip with FPV for a mowing experience from the future

Can Autonomous and RC mowers co-exist?

OxChief’s goal is bringing autonomous mowing to every mower. When we released the OxChief software along with the OxChief Alpha servo control for the Bad Boy Maverick HD, we were happy to offer the first mowing autopilot solution designed for anyone with a large (think multi-acre) mowing area.

While the future of autonomous mowing is bright, RC mowing is very useful in many applications.

RC mowing with an FPV setup, in particular, may just be a unique golden thread connecting the fun of being a boy with the need for productivity of being a man.

I just bought a Bad Boy mower at Tractor Supply — how long does OxChief install take?

One Afternoon.

If you presently own a Bad Boy mower, just pick up an OxChief RC control and you’re ready to roll.

Otherwise, head over to your local Tractor Supply (or, even better, your local Bad Boy dealer) and pick up your new mower.

Having secured your OxChief RC control and your mower, let’s take a quick overview of the process of installing the RC control onto your mower

RC Control on Mower Install Executive Summary

Here are the steps:

RC Control on Mower Install Detail

Enjoy your complete zero-to-hero tutorial/demo covering the entire install:

Hope you’ll head over to the store and take the plunge.

You’re going to love it.

–Wayne

We’re excited to introduce OxChief.

OxChief’s goal is to make driving your mower optional.

OxChief provides software and hardware to make this possible.

https://github.com/oxchief/oxchief-client

https://shop.oxchief.com/

Software written and hardware manufactured in the USA.

What I thought would take a year ended up being most of a decade.

Here’s hoping you enjoy.

Exhibit A:

That’s a timelapse of a Scag (the dominant heavy-duty commercial mower brand in my corner of the world) 72″ Turf Tiger skid-steer mower mowing over some old gentle farm terraces next to our grass runway. The mower is running at approximately 3.8 miles per hour (1.7 meters per second).

This run (and many more) was achieved without any human intervention — but you’ll notice that I follow it carefully and monitor the behavior as this was an early run with the particular build / configuration.

Lofty Goal & Cordial Invitation to You

The goal of this article is to clearly present the components of that mower for your consideration. As usual, if I forget something (as is usually always the case) — please drop a line in the comments and I’ll try to color in whatever seems vague. I really do appreciate all the comments and questions that readers have submitted on these posts over the last couple of years — when I first made the leap from a software-centric world to robotics world there were so many fire hydrants of information from so many engineering disciplines that it felt like I was underwater for years. Humans aren’t born possessing much of even the basic knowledge that underpins all of robotics — most eager novice robot builders first encountering 5v, GND, VCC sockets on an expensive board will be massively nervous about committing to what they assume those letters mean — and perhaps they’ll utter a silent prayer hoping that the board doesn’t smoke and simmer when they make the leap and plug in their breadboard wires. In other words: if you have a question about the project (even if it seems much more “elementary” than whatever other comments are being made), please don’t hesitate to log the question — you can just use a throwaway account if you want to wait until you’re some kind of robot ninja to disclose your identity to the world.

How the Build Went Down

The electronic footprint we added to the big Scag automower is very similar to the smaller automowing rig we’ve previously discussed. The points of dissimilarity between these mowers proved to be some of the places that required the most creativity — most notably being selecting and implementing the servos (i.e. the motor/shaft apparatus that pushes the mower handlebars back and forth).

The Quest for the Perfect Servo (or “An Ode to Pops and Wingxine”)

You are likely familiar with skid-steer mowers, but if not, here’s what happens — you’ve got 2 handles — one for your right hand and the other for your left hand — when you push the right handle forward of neutral, it makes the right wheel go forward (the further you push it forward, the faster the wheel goes) — when you pull it back behind neutral it makes the right wheel go backwards (the further you pull it back the faster the wheel goes). Same logic for the left handle and the left wheel. So, the idea of a skid-steer vehicle is that you’ve got 2 inputs that have an identical bearing on both speed and direction (as opposed to a traditional vehicle where 1 input (the gas pedal) has the biggest impact on speed and the other input (the steering wheel) largely determines direction).

Our challenge is to come up with some mechanical component to attach to either handle and then to configure that component to work nicely with our autopilot.

I have one chief requirement for the servos that control the drive arms: the servos must be overridable by an average-strength person in any reasonably imaginable scenario.

Pioneering Efforts

Way back in 2005 a couple of highly talented engineers published a paper giving an overview of how they had equipped a Scag Turf Tiger with radio control in support of government rollover testing.

Circa 2005 Scag radio control conversion

The craftsmanship those guys showed in that conversion was impressive, and their design apparently met the government rollover testing needs perfectly — but they chose linear actuators to move the drive arms and this is not appropriate for robots that are operating in the presence of humans^[1].

Good luck overriding these beasts.

Linear actuators typically have a high static (holding) load rating — so when you cut power to the actuator (or attempt to override it when it is powered) the actuator is incredibly difficult to override. Working around 1500 pound blade-slinging autonomous machines is stressful enough without the thought that I can’t manually override the control arms.

But if we just throw out linear actuators on safety grounds (assuming your robot will ever be around people) and convenience/practicality grounds (I want to effortlessly drive the big Scag around with the autopilot turned off without unhooking the servos from the control arms), what other option do we have for controlling the arms?

In order to answer that question, I’ll need introduce you to two folks: Wingxine and Pops.

Who is Wingxine?

The identity of Wingxine is largely shrouded in mystery, but this enigmatic genius in the Far East has secured entry into the Huge Skid-Steer Robot Hall of Fame by making and then delivering upon the following observation:

Wingxine’s moment of robotics glory arrived as she or he realized that if you mount an H Bridge motor driver on a cheap-reliable-torquey-proven-ubiquitous TAKANAWA 555 Metal Gear Motor that you’ve just minted some legit servo Graphene from readily available components.

If you spend much time sourcing robot parts from folks on the other side of the globe, you’ll eventually run across a few who simply stand out in terms of the quality of components they consistently create and deliver. Wingxine is precisely that kind of individual: you are strongly advised to check out his/her eBay presence and AliExpress store.

Mower Control Arm Servo Overview

The specific servos we’re using to drive the big Scag mower are two Wingxine ASMC-03A servos:

The Wingxine servos are plenty powerful (even when powered at 12 volts) to capably actuate a zero-turn mower’s drive arms while not so strong that they can’t be overridden by an adrenaline-charged operator in a pinch. As an uber-welcome bonus, when powered off they don’t interfere with normal mower operation.

Note that the ASMC-03 servo is now deprecated in favor of the newer ASMC-04, and you can pick it up in either the ASMC-04A or ASMC-04B flavor.

Seeing that the ASMC-03A is deprecated, I went ahead and attached all the stock documentation above (largely for selfish reasons) because this stuff sometimes has a way of vanishing from the internet^[2] (and we still have a dozen or so of these servos laying around the shop).

Since the ASMC-04A is the standard going forward, let’s go ahead and grab those docs too:

The long answer is that you may need the 04B for your application (or another similar servo from Wingxine’s diverse selection), but the 04A is precisely what we need for this application.

Here’s the deal: the 04A moves 4.17x faster than the 04B at the cost of torque — 39% less torque. Trust me though — this servo has A LOT more torque than you may imagine if you haven’t ever worked with one of these proven Takanawa 555 gear motors.

Here are the official performance specs (at 24v) for both servos:

ASMC-04A is rated at 60 degrees of rotation in 0.12 seconds -- with torque of 110kg.cm.

ASMC-04B is rated at 60 degrees of rotation in 0.50 seconds -- with torque of 180kg.cm.

Important caveat — those numbers are when the servos are powered at 24v, but I run the servos at 12v because 12v provides PLENTY of power to quickly push the control arms. I haven’t seen official numbers for 12v, but I’d ballpark guess you get half the speed/torque^[3].

So, for safety, the servo’s ability to pull back the control arms (thus stopping the big mower) in the blink of an eye is very important — an extra second could be the difference between a close call and disaster.

Connecting a Mighty Wingxine Servo to a Zero-Turn Mower

As a single guy I used to marvel at how much easier it is to buy the world’s fastest superbike than to get a date with your crush (look, if you’re born with the physical gifts of ‘ole Matt then maybe getting the date is actually easier — I wouldn’t know).

Through persistence (and with the help of your ’06 GSX-R1000), you may eventually get the date.

And, of course, if all goes well, the date may turn into several dates which leads to an entirely different kind of ceremony.

Next you’re in for a real ride because a little fellow (or little lady) will likely appear who occupies such a huge place in your heart that the bike’s value, in comparison, seems negligible.

Yes, I know, this is an engineering blog and you strapping young engineer won’t believe it unless it’s in a formula.  Here you go:

You think your love for that awesome toy or hobby can’t be superseded? Just wait

You end up selling the superbike for fear that it will cause your kids to become orphans, and, in an unpredictable pivot that you never imagined, you suddenly become interested in the hitherto-numbingly-boring hobbies of your forebears, such as growing absurdly big tomato plants^[4].

And so, my young engineer brother, by all means take your bike for a spin^[5], and somehow find the courage to ask out the girl — and remember that your decisions are rarely idempotent.

Weren’t we talking about servos?

OK, back to business. So the point of our little digression above is clearly that the easy part is buying the ASMC-04A servo from Wingxine — but the hard part is figuring out how to control your zero-turn mower’s arm with that servo.

Fear not — while I can’t tell you how you’re supposed to ask out the girl (but I do think that old-school ask-her-out-face-to-face [and often dealing with rejection on the spot] is more manly than all this Snapchat jazz the kiddos like to play these days), I can show you how we hooked up the servos.

Servo Linkage Overview

Let’s jump right to a load of photos (with scattered comments) of the servo-to-mower linkage before digging into the details:

Removing the Control Arm Dampers and Reverse Springs

Our Scag has a damper on each control arm that will degrade the servo’s performance if left attached. Additionally, pulling each control arm behind neutral (i.e. reverse) depresses a metal resistance spring that adds significant torque to the effort required to move the handles throughout the reverse range of motion. When operating at 12v the Wingxine servo lacks sufficient torque to power through these springs, so we’ll remove them as well.

A careful look near the end of each damper reveals a tiny wire around the neck that secures the damper to the mower — you’ll need to pivot the free end of this wire up and then pull the wire out.

Now that the mower’s control arm dampers and the reverse-resistance-springs are removed, it’s time to build the servo linkages that will move the mower’s control arms. Let’s take a look at the parts you’ll need to make it happen (i.e. to build a set of left and right servo linkages).

Servo Linkage Parts List

1. 2x Wingxine servo
2. 2x Wingxine servo plate (at the time of this writing it’s 10 bucks on AliExpress but 20USD on eBay)
3. 4x male ¼” rod end (fee free to substitute Metric components — regrettably somehow we ended up SAE for these parts).
4. 4x ¼” x 1⅜” long hex cap bolts — you thread these bolts through the rod end into the servo plate to connect the rod end to the servo plate. You’ll also thread the bolts through the other rod end to connect that rod end to the metal “tab” you’ll cut out and weld to the mower’s drive arm linkage.
5. 6x ¼-20 hex nuts — on each servo assembly you’ll use 2 nuts to secure the servo plate to the rod end and 1 nut at the other end of our custom 12″ servo-to-drive-arm-linkage to connect that rod end to the ⅜” steel tab we’ll weld to the mower’s drive arm linkage. 
6. 2x rectangular sections of ³⁄₁₆ steel plate — you cut the plate roughly 6″ wide x 9″ long for the left/right brackets that we mount to the mower as the base for the servo. You’ll note in the pictures that we put a 90° bend (achieved with a press brake) in the plate at the 2.5 inch mark. Keen eyes may observe that we actually used ⅛ steel plate — this was an error. Our brackets flex slightly during rapid steering introducing a little imprecision into the robot’s performance.
7. 8x M4-0.7 12mm bolts to connect the servos (4 for each servo) to the custom brackets we fabricate.
8. Washer assortment for use when bolting servo to bracket
9. Threadlocker to use primarily on bolts that tighten the servo plate to the servo shaft and optionally on bolts that connect servo to mounting bracket.
10. 2x 12″ long by ½ thick steel tube. We’re going to weld a nut to either end of this tubing to create the adjustable linkage shaft between the servo and the mower’s drive-arm linkage (I realize that that sentence likely makes sense to no-one except the author, but I hope that after looking at the pictures you can see what we’re using that steel tube for — note in the pictures that the steel tube linkage is painted black).
11. 8x 1/4-28 nuts. We’re going to weld one of these nuts to either end of the 12″ steel tubes (note that you really can’t see these 4 nuts in the pictures — it just appears that our 12″ black tubes are threaded). We’ll then screw the male rod ends into these nuts and we’ll use this adjustability (i.e. by screwing the rod end into the nut we make the linkage shorter and by unscrewing it we make the linkage longer — this is where selecting fine threading on the rod ends pays off because it allows us to really fine tune this adjustment) to get the servos and the mower drive arms synced up very nicely. The other 4 nuts are used to tighten the male rod ends to our 12″ linkage once we’ve got the servo-to-drive-arm adjustment nailed down.
12. ⅜ thick x 2″ wide steel bar like this to cut out 2 teardrop or rectangular “tabs” that we’ll weld to the mower’s drive-arm linkages.
13. 3 heavy duty bolts (such as these M10 – 1.50 x 20mm) and nuts (such as these) to secure the right servo bracket to the mower. Note that on our mower the left bracket happens to be lined up exactly opposite from 2 substantial bolts that connect a drive-belt pulley to the mower, so we just piggy-back off those 2 bolts to connect the left bracket to the mower. 
14. 1x M4-0.70 x 8mm long (or shorter) hex screw to secure the servo plate against the servo shaft.
15. 1x M5-0.80 x 10mm long (or shorter) hex screw to secure the servo plate against the servo shaft.
16. 1x can of Rust-Oleum flat black paint (or whatever brand/color suits your fancy).

That should be it for the linkage parts. Hopefully when you review the pictures of the linkage above you’ll get a feel for what’s going on.

Some additional observations may prove useful if you’re planning on recreating a setup similar to this.

Preparing a Wingxine Servo for it’s Ultimate Life Mission

There are a few distinct steps in the wingxine_servo-to-servo_plate-to-linkage configuration dance. We have to start somewhere, so let’s start here: you’ll need to come up with some way of sending PWM values to your Wingxine servo.

Manually Controlling the Wingxine Servo From Within Mission Planner

A slick (and perhaps under-appreciated) feature of Mission Planner is it’s ability to control ArduPilot via a USB joystick connected to your PC. And so, for this exercise, let’s assume you have the following components:

1. A Flight Controller such as the hot-promising-and-inexpensive Kakute F7 (of course your fav flavor of Pixhawk should work as well) (also note that if you’re unboxing a fresh Kakute F7, you’ll need to follow these directions (just one time) to replace the firmware shipped on your Kakute with ArduPilot. The file you’ll load for rover is the ardurover_with_bl.hex file found here).
2. A USB Joystick (don’t cheap out below this price point bro unless times are really tough).
3. A 12v power source (such as a barely-breathing red lead-acid battery salvaged from yet another electric wheelchair I recently brought home to the rolling-but-understanding-and-beautiful eyes of the wife) and some way to connect it to the Wingxine servos (such the aligator clips and JST connectors I use below).
4. Some random wire to connect Servo Output 1 from your Flight Controller to the Signal pin of your Wingxine servo.
5. Another random wire to common up the ground between the Wingxine servo and your flight controller — in other words (i.e. if you’re new to electronics world) suppose your flight controller is powered by your PC via USB, but the Wingxine servo is powered by a 12v battery. You achieve this “common up the ground” idea by connecting the Ground/GND/G pin from the Wingxine servo to a Ground pin on your Flight Controller. Guess what happens if you forget this step? Usually nothing
6. Of course you’ll need Mission Planner and your choice of Wingxine servo and Wingxine servo plate (links above).

You’re going to hook all of those components together so that you get something like this:

Wingxine servo connected to Kakute F7. The F7 is powered via USB from the the Mac and the Wingxine servo is powered via the big red 12v battery in the background. Just remember that since we’re powering these devices from different sources that we’ll need to wire the F7’s GND to the Wingxine servo’s GND. Note the joystick’s USB is also connected to the Mac — Mac is running Mission Planner via Parallels VM.

If you’re using a Kakute F7, then it may not be obvious to you (as it was not to me) which wires in the F7’s wiring plug correspond to ArduPilot Servo Output 1 and Servo Output 3 (i.e. as a fun little twist, Kakute actually calls the Servo Output 3 wire “M4” and the Servo Output 1 wire “M2”). Here’s a little picture I made to hopefully clear that up:

Kakute F7 wires corresponding to ArduPilot Servo 1 and Servo 3.

Assuming you’re rolling a wingxine_servo-flight_controller-joystick-mission_planner setup similar to the pictured above, let’s talk about the ArduPilot configuration we’ll set from within Mission Planner so that we can send PWM values to the Wingxine servo.

Mission Planner Joystick Configuration Steps

1. First off, go to Config/Tuning –> Full Parameter Tree –> SERVO1 and set the values like these:
   

   SERVO1_MAX is the PWM value that ArduPilot will send to the Wingxine servo when we push the joystick all the way forward. SERVO_1_MIN will correspond to the joystick pulled all the way back. SERVO1_TRIM is the PWM value ArduPilot sends to the servo at rest — we’ll use this value later on to dial in the exact PWM value that moves the mower’s handle to the neutral position.

2. Now find the PILOT_STEER_TYPE parameter and set it like so:
   

   Note that we’re using the filter over in the lower right corner to hone-in on the PILOT_STEER_TYPE param

3. From the Mission Planner main screen, select the “Actions” tab in the lower-left-side tab group and click the “Joystick” button:
4. From the Joystick popup window, click the “Enable” button at the top, ensure that the dropdown box next to “Ch 1” (near the top) has “Y” selected, and check the “Reverse” box to the right of “Ch 1”, i.e.:
   

   Don’t forget to click the “Save” button near the top before exiting this screen.

Now assuming that all goes well, you should see the green bar next to “Ch 1” grow when you push the joystick forward and it should shrink when you pull the joystick back. When the joystick is at rest the bar should be in the middle.

1. Set the flight controller to Manual mode (for joystick control) like this: 
2. Arm the flight controller by clicking the “Arm/Disarm” button, i.e.: 

Now on the Mission Planner main screen, click the “Status” tab in the lower left tab box group and scroll over until you see the value of “ch1out”. Now when you push the joystick forward and backward you should see the “ch1out” value move between 2000 and 1000, i.e.:

If your flight controller’s Servo1 output pin is hooked up to the Wingxine servo’s signal pin (i.e. like the setup in the big picture with the red 12v battery, the servo, et al earlier), then you should now be able to turn the servo plate by actuating the joystick.

Dialing-back the Wingxine Servo to the Maximum Safe Range

Now that you’re controlling your Wingxine servo with a joystick connected to Mission Planner, it’s time to scale back the servo’s range so that it’s approximately 150° (maximum) — much more on the reasoning for this in a minute. Let’s go ahead and drop a video showing you how to make that adjustment and prepping you for the discussion ahead:

If you watched that video, did you notice how, by good fortune, the travel of the Scag right servo linkage is almost exactly 33 millimeters? So for the Scag, using the holes in the servo plate provide exactly the amount of horizontal travel needed to move the mower handle throughout it’s entire range of operation. Note that, even though we have the full range available, on our current setup with the big Scag we’re limiting the forward throttle to approximately 60%. The reason for this is that, in the event of some unexpected failure, I’m not yet wanting the Scag barreling down on someone at 10 MPH.

Now that we’ve thrown around this 33-millimeters-of-travel number a few times, it’s probably appropriate to make another corollary observation: if your setup needs more than 33 millimeters of travel, then you don’t need to use the holes on the Wingxine servo plate. You’re going to have to put on your engineer cap and figure out a way to extend off that plate a little so that your control arm doesn’t end up pulling the Wingxine servo into a parallel-with-the-servo-shaft nightmare scenario.

Since we’ve broached the nightmare scenario topic, let’s take a moment to dig in on that a little more deeply before proceeding.

A Soothsayer Bids You Beware 3 O’Clock and 9 O’Clock

Remember that we’ve got 1 cardinal rule for the servos — they must be overridable by an average-strength person in any reasonably imaginable scenario. And remember that I’ve postulated that the Wingxine servos and the linkage we’re employing are overridable — but there’s one big land-mine that you must be aware of lest your mower become HORRIFICALLY STUCK IN FULL-SPEED FORWARD OR FULL-SPEED REVERSE.

We’ll get to some pictures of the nightmare scenario in a bit, but first take a few seconds to peruse the collection of photos below showing our servo linkage configured in a manner that should be easily overridable.

Here is the right-side servo linkage at full throttle forward and full throttle reverse:

Here we show the left-side servo linkage at full throttle forward and full throttle reverse:

The full throttle forward and reverse linkage pics above reveal that the servo linkage stays safely below the parallel line of the servo shaft (shown by the bottom edge of the black square) throughout the entire range of motion. In other words, the force vector of the linkage is never entirely directed into the servo shaft — we always want there to be a component of the linkage’s force vector that’s pointed down (or up, depending on your configuration — the point is you never want to be in a place where pulling or pushing the mower handles sends 100% force directly into the servo shaft) so that our pushing or pulling the mower handles will result in the servo rotating (well, here again this description seems clunky, but hopefully the pictures below show you what to AVOID).



Here are several different angles of these two linkage positions you must ensure CAN NEVER happen (either by the user actuating the mower handles or by the servo rotating to some demanded position):

If your servos can get to points where the force vector of the servo linkage is directly inline with the servo shaft, you are courting disaster.

If anyone reading this is actually planning on using the design above, I would humbly ask the following: as an exercise, please turn on your Wingxine servo, bump back the shaft angle range so that it can rotate up to 3 o’clock or 9 o’clock, and give it whatever PWM is necessary to get to those 3 o’clock / 9 o’clock scenarios on your build — so just to be clear, now you have the Wingxine servo turned on and it’s holding the linkage parallel to the center of the servo. Now go ahead and try to move the mower arm back to neutral — not so fun eh? Imagine trying to move those handles out of that position in some unthinkable scenario when you’ve got no seconds to spare.

Fabricating the Servo Linkage Parts

There are 3 pieces to the linkage puzzle that you’ll need to fabricate:

1. You’ll precision weld a 1/4-28 fine threaded nut (suggested in the parts list earlier) to either end of the 12″ steel rod (also in the parts list) to create the adjustable linkage shaft
2. You’ll cut out the 6″x9″ servo bracket from the ³⁄₁₆ plate in the parts list and put a 90° bend in the plate at the 2.5″ mark
3. You’ll cut out the linkage connecting tab from the ⅜ steel bar in the parts list, drill a hole near the top, and weld the tab to the mower’s control arm

Now is a good time to be friends with a talented welder. If your welding buddy is not returning your calls, then perhaps just buy a solid aluminum rod and tap threads into either end of the rod.

Use press brake to put 90° bend in 6×9″ ³⁄₁₆ plate at the 2.5″ mark — remember that your bracket should appear thicker than this picture unless you’re planning on reinforcing the bracket. Unless you specialize in blindfolded chess or similar feats of precision imagination, I’d recommend building your Wingxine servo / bracket assembly before deciding which hole in the Wingxine servo plate to use for bolting on the male rod end (in other words, depending on which way you end up mounting your servo to the bracket, what you think is BDC may actually be TDC) .

Connecting “Tab” Considerations

If I haven’t succeeded in conveying the parts / methods needed to pull off this design, just ping me down in the comments and I think we’ll be able to get you in a good spot.

Mounting the Linkage Onto the Mower

Bolting the Wingxine Servos to the Servo Brackets

Now it’s time to bolt the Wingxine servos to our custom servo brackets. You’ll need to drill out 4 holes roughly in the center bracket in a rectangular pattern matching the 4 holes on the back of the wingxine servo. I’d suggest using a washer or 2 and some Loctite for each of these M4-0.7mm bolts so that you’re not tempted to over-tighten the bolt and strip out the thread.

Mounting the Servo Brackets to the Mower Frame

For the right-hand bracket, we were free from obstructions. After determining a location where the rod end that’s bolted to the servo plate will be flush above the mower control arm, we drilled 3 holes in the bracket and 3 holes in the mower and bolted the bracket to the mower.

Right side of the Scag was clear of obstructions. We determine the location to mount the servo bracket where the rod end connected to the servo plate will be approximately inline with the mower’s control arm. In other words, we want the servo linkage to be directly above the mower control arm. On our mower we mounted the right bracket so that the rod end connecting the linkage to the servo plate is ~3″ to 3.5″ (the precise distance that this rod end is above the mower arm fluctuates as the servo plate rotates) above the Scag’s mower arm directly below. Note that the other end of the connection (i.e. where the other rod end is bolted to the welded tab) is approximately 1.5″ above the control arm.

On the left side two mounting bolts for a tensioner pulley were exactly where we needed to be, so we just drilled 2 holes into the bracket to line up with the pulley’s mounting bolts and dropped our bracket right in place. Note we had to cut out a lobe from the center of the bracket to clear the pulley’s center bolt.

Tensioner pulley bolts and nuts were exactly where we wanted to mount the bracket — fortunately the bolts had plenty of length available to sandwich in the servo bracket.

To be fair — the preceding pictures paint a cute picture of one hand dropping in the servo bracket and then it magically finding itself bolted in right here with the tensioner pulley nuts. If you’re familiar with tensioner pulleys, you know that whole operation was a slight (but not terrible) headache. Note that the left servo sits about 1″ closer to the mower control arm (visible in the picture above as a metal bar running left-to-right directly below the servo plate) than the right servo.

The exposed PCB on the servo enables quick adjustment (in particular the servo shaft’s range is controlled by a potentiometer — this is the inner-most blue plastic shroud visible atop the PCB in the picture directly above) but it’s clearly a temporary solution — it’s entirely reasonable to expect longevity out of these servos if we properly shield them from the elements. For now we’re going to leave that exercise to some future blog post if this discussion generates enough interest to warrant revisiting this setup.

Securing the Servo Plate to the Servo

If the tightening bolts shipped with the servo plate interfere with your connection just replace the obstructing bolt with the appropriate grub/allen/set screw we listed in the big parts list earlier.

Bolting the Male Rod End to the Servo Plate

Male rod end connected to servo plate with a ¼ x 1⅜” long hex cap bolt and two ¼-20 hex nuts.

Now go ahead and thread the 12″ servo linkage we manufactured about half-way into the male rod end pictured above.

The nut on the male rod end above is set at about the half-way point on that rod end.

Welding the Tab to the Mower Control Arm and Bolting the Rod End to the Tab

Now that the servo bracket is mounted to the mower and the linkage is completely assembled (and the adjustable points on the linkage are around their mid-point), it’s time to bolt the connecting tab to the rod end at the other end of the adjustable linkage and mark the points on the mower’s control shaft where our tab is resting. Now go ahead and weld the tab to the control shaft and bolt the rod end to the tab, i.e.:

Connecting tab welded to mower control shaft. Rod end connected to tab via ¼ x 1⅜” long hex cap bolt and ¼-20 hex nut. Any future reader who would like to submit a functionally equivalent uniformly metric parts list (along with links for purchase) for the linkage assembly is greatly encouraged to post the fruits of that endeavor.

Adjusting the Servo Linkage so that Full Range of Motion Stays Safely Below the 3 O’Clock – 9 O’Clock Line of Disaster

Unbolt the rod end, loosen the nut that tightens the rod end to the linkage shaft, and turn the rod end to increase or decrease the length of the servo shaft.

All the warnings shouted from rooftops above about the 3 o’clock and 9 o’clock positions of the servo plate will hopefully now encourage you to move the servo arm throughout it’s range of motion and note the entire travel path of the servo plate. Assuming that your servo plate connection provides sufficient horizontal travel, you should be able to adjust the linkage to stay in the safe zone. To me it’s easiest to adjust the servo linkage at the welded tab end (only 1 nut to loosen to disconnect the rod end at that end). Remember that our rod ends are identically threaded, so you will have to unbolt the rod end from the welded tab (or unbolt the other rod end from the servo plate), loosen off the tightening nut securing the rod end against the servo linkage, and spin the rod end to increase or decrease the length of the servo linkage.

Once you’ve got the servo linkage adjusted to a length where moving the mower’s handle full forward and full reverse translate to a nice safe arc back at the servo plate connection, you may want to give yourself a good pat on the back — and then go ahead and give all the linkage nuts a quick check to ensure they’re all tightened.

Now we’re finished mechanically connecting both servo assemblies to the mower.

It’s time to give them juice.

Wiring up the Servos

You’ll observe 3 wires running to either Wingxine servo — 12+, GND, and PWM signal from the Pixhawk.

Note that we supply the 12v (fused at 5 amps in the electric box) through a 10 amp diode. The reason is that the servo motor becomes a generator if you rotate the shaft when the motor is turned off. The diode prevents the servo from feeding electricity back into our system.

That yellow wire snaking in from the upper-right corner is the PWM input to the board from the Pixhawk. Note that we drop a big gob of hot glue over this connection and all the jumpers to keep them from wiggling loose over time.

We’ll be able to trace the journey those PWM wires make from the servo up to the electronics box and then back down to the Pixhawk (mounted beneath the footrest on the Scag) a little later.

Configuring the Servos for ArduPilot in Mission Planner

This ArduPilot configuration dance is really just figuring out what PWM value makes each Wingxine servo go to neutral and then which PWM values tell the Wingxine servos to move to the maximum forward and then the maximum reverse position that you’re comfortable with. Restraint is obviously encouraged here — especially on early runs with a new build.

On skid-steer vehicles, ArduPilot controls the left wheel via the “Servo1” settings and the right wheel via the “Servo3” settings.

Random picture snapped when I was figuring out SERVO3_TRIM (remember that the “_TRIM” value is the PWM value that moves your mower’s arm to the Neutral position), SERVO3_MIN, and SERVO3_MAX ArduPilot values for Servo3 (the right hand servo). During this calibration, the servo is controlled by the USB joystick on the right — see instructions earlier in this post for details on connecting joystick to Mission Planner and sending exact PWM values to servo.

Servo1 (left servo) values for our Scag — note that your MAX/MIN/TRIM values will likely differ at least slightly from these.

Servo3 (right servo) values we ended up with on the big Scag — note that we tell ArduPilot that the lower bound of the PWM range corresponds to forward on this servo via the SERVO3_REVERSED setting. If you’re using the newer ASMC-04 servos and have forward corresponding to the upper PWM range (i.e. sending 2000 PWM to both servos pushes both left and right mower arms forward), then you won’t need to fiddle with SERVO3_REVERSED.

Once you’ve invested a little time moving both mower arms to full reverse, full forward and neutral with the joystick, you should be able to ballpark those 3 values for each arm on your robot. Just remember that the PWM values in Mission Planner you’re actually sending to the servos (and thus where you’ll find the values you’re going to enter in the SERVO1/SERVO3 settings) are ch1out and ch3out — it’s easy for your eyes to lose track and end up looking at the ch1in and ch3in numbers.

Mission Planner gives the PWM values we’re sending to our servos in the “Status” tab in the lower left hand corner. Remember when you’re actuating the servos with your Joystick that the PWM value going to the right servo is ch3out — the value going to the left servo is ch1out.

This servo setup has been a fair amount of effort, but eventually the setup will start to function similar to this:

Hopefully we’ve covered enough to at least paint a picture in your mind of the points involved in this setup.

If you’re following along with a similar setup but something isn’t quite clicking from the ArduPilot side, then maybe download your configuration params in Mission Planner and compare them with the params on our build.

Electronics

The main electronics box (pictured above) contains all the components that aren’t picky about their mounting location. The cast of characters here is largely the same as we’ve seen in previously discussed builds with the exception that the Pixhawk is not mounted in this box — it’s down below closer to the rover’s center of gravity (more on this in a bit). You’ve likely noticed that the wiring in the box could use some spring cleaning — somehow I’m pretty zealous about labelling but lazy about tidying up the wires too early.

We use a standard electronics box and give it a coat of silver paint. Toggle switch is the main power switch and button arms ArduPilot. Note the power, telemetry radio antenna, and compass wires on the front side of the box.

Metal antenna cable terminal attached to the rear right side of the electronics box is the GNSS antenna cable. Emerging from the rear of the box is the long USB cable (which travels from the box, down through the conduit hose, and into the flight controller enclosure beneath the floorpan), a telemetry radio antenna cable, and the power/signal wires to the right servo.

The base for the main electronics box is an aluminum plate riveted to a steel bracket. Aluminum is much easier on your drill-bits as you prospect for that perfect hole location.

Conduit hose (left-most hose pictured here) provides a robust waterproof connection between the main electronics box and the aluminum flight controller enclosure beneath the floorpan.

Conduit sits between the back of the seat on the front of the radiator — in retrospect I’d opt for larger conduit to make tunneling the wires through it less challenging (this is 1/2″ inner diameter). Note the main power +- wires and the left servo wires (inside the black plastic shield) are tied to the conduit.

Conduit travels down until reaching the custom aluminum flight controller enclosure beneath the mowing deck and the floorboard.

Aluminum enclosure location is close to vehicle center of gravity to help the accelerometer and gyro performance. Thanks to Pops for the slick enclosure design and fab.

Conduit connection to flight controller enclosure face plate. Note that an old original Pixhawk is pictured here, but we’ve switched over to a Pixhawk 4 at publish time.

Flight controller sits on a velcro-covered plate.

Pixhawk 4 wired up and ready to place on velcro pad in flight controller.

A little extra vibration dampening never seems to hurt.

White velcro pads taped to the flight controller’s base mate to the aluminum enclosure’s velcro base.

You may have noticed the white braided string in these pictures (visible riding atop the green compass wire above) — this string serves as a guide when running new wires through the conduit. So, just to be clear, if you want to run a new wire, you securely tape the end of the wire to the string and then just pull the string from the other end.

You’ll need a fair length of excess string on either end to pull off the string-wire-guide maneuver — I just wind the excess string around a golf tee when finished running a wire.

The silver rectangle peeking out beneath the floorpan is the front of the flight controller enclosure.

Additional Pictures for (Hopeful) Clarity

Let’s go ahead and drop a collection of pics taken at different angles to hopefully clarify pieces we may have overlooked.

Aluminum Roll Bar

The electronic components mounted atop the aluminum roll bar don’t want to break up, but they do need some space.

Moving left-to-right we’ve got a multi-frequency GNSS antenna, a strobe light (especially important to run this when the mower is working on the airplane runway), the silver compass box flanked on either side by telemetry radio antennas, and the short black telemetry radio antenna on the far right. Why do we have three telemetry antennas you may ask? I like to have two telemetry connections available for communication with the flight controller and the RTK corrections get their own dedicated radio.

Compass Enclosure Insulation

At this point I’m sure you’re on the edge of your seat to find out what kind of wizardry is happening within the silver compass enclosure in the center of the roll bar. Well, my friend, your wait is almost over.

The wrinkled silver lid is just a piece of aluminum foil hot glued to a piece of cardboard. The cardboard is attached to the clear plastic lid through the wonders of Velcro.

Well this is slightly embarrassing — the enclosure is stuffed tight with an old pair of Fruit of the Looms and several old socks. The selection of cotton cloths may make your author blush, but the performance of our well-insulated compass beneath those tighties is nothing to be ashamed of.

Pulling everything back we finally strike gold and find a tried and true Honeywell 5883l compass that, of course, is hot-glued to the base of the enclosure.

Since we’ve now broached the sensitive subject of compasses (or “magnetometers” / “mags” if that suits your fancy), let’s go ahead and cover one last important piece of information about this build — we’ve made a minor change to the ArduPilot source to enable fine-tuning the compass. Don’t worry, I think you’ll find that compiling & rolling your own changes to ArduPilot is a relatively painless operation.

ArduPilot Fork with Compass Offset Configuration

Inexpensive, readily-available (albeit now discontinued) Honeywell HMC5883L compasses can provide solid heading readings at high frequency, but you have to keep a few considerations in mind:

1. They don’t respond well to sudden temperature fluctuations (i.e. you need to insulate your compass from sudden temperature changes (think sun suddenly pops out from behind the clouds)) — on the big Scag I’ve just stuffed the compass box full of random cloths (as insulation) and stuck some aluminum foil on top to mitigate the effects of a sudden external temperature shift often caused by the the sun’s appearance. The more standard (and committal) way of insulating the compass would involve entombing the little board (after you’ve soldered wires to communicate with it and said your last goodbyes) in some thermally insulating electronic potting compound — like this cheap stuff that most people buy or the MG Chemicals stuff you buy if your customer needs a little more traceability. Another hack if you like hacks is to mount the little fella in the enclosure, wire him up, and then pump the enclosure full of hot glue (recommended if you’re one of us who find squeezing out big globs of hot glue to be oddly therapeutic — also recommended if you’re trying to wean off a Dr. Pimple Popper addiction that the wife finds repulsive).
2. They’re very sensitive to iron (of course) and this may sneak up on you — for instance, if you get within a couple of meters of a wire fence or a metal gate it can throw off your heading reading by several degrees.
3. Their notions of North / South / East and West all need calibration beyond what is presently available in Ardupilot. I take care of this by enabling you to enter N-S-E-W offsets in the Ardupilot codebase fork we discuss below.

Compiling Ardupilot Source Does Not Require a PhD

So I’ve forked the official ArduPilot source and added a little bit of code to allow the user to set offsets for North/South/East/West.

Great, you say, but no-one wants to spend 2 days figuring out how to compile and load some custom autopilot code. Agreed — and fortunately compiling this code and loading it onto your Pixhawk4 / (other flight controller) is a pretty straightforward procedure.

If you’re using a Mac, this is the entire set of commands (assuming 1. you have git installed and 2. you have your flight controller connected via USB to your Mac and 3. your flight controller is Pixhawk4 (if it’s some other board, I’ll cover you in one second)):

git clone https://github.com/waynebaswell/ardupilot.git ardupilot-baswell
cd ardupilot-baswell
git submodule update --init --recursive
./waf configure --board Pixhawk4
./waf rover
./waf --target bin/ardurover --upload

If you’re running some board other than Pixhawk4, then run this command (from the ardupilot-baswell directory) to print waf configure help:

./waf configure -h

Scroll down through that big beautiful nicely-formatted help text until you see help for the --board option, i.e.:

In a world full of caveats, alas ArduPilot builds are no exception. Recently ArduPilot has moved to a new base firmware called ChibiOS (that, among other advantages, greatly simplifies the code required to port ArduPilot to new boards).

And so, if instead of Pixhawk4 you’re running Pixhawk2/Cube (please proceed with caution as I haven’t tested this code on a Pixhawk2), you may be tempted to use one of the Cube.. board options above — but those options are actually supporting the old base firmware.

A quick review of the highly informative official ArduPilot build instructions indicates that we’d use the following waf command to configure Pixhawk2/Cube using ChibiOS:

./waf configure --board fmuv3

Fine-Tuning the Compass

OK, now that we’ve got this custom ardupilot fork loaded, how are we going to use it to fine tune our compass? Here’s how it plays out — go outside with your robot and point it due North. Now assume you’re quite certain that the robot is pointed due North (0°) but Ardupilot is telling you you’re at 1°. So our ArduPilot fork gives you a way to go into Mission Planner and set a parameter to indicate that the compass’s notion of North is actually 1° wrong. OK, make a note of that 1 degree — we’ll use it in a bit. Now pivot your robot so that it’s pointing due East (90°) — suppose that ArduPilot is indicating 96° here — make a note of that too. So you point due South (180°) and Mission Planner indicates 174° — note this. And lastly you aim due West (270°), but Mission Planner proudly indicates 266.5° — record this as well.

Note that on our big Scag, when we engage the blades it has a slight effect on the compass (even though the compass is almost 2 meters away from the blades). In order to get the heading as close to true as possible, we always keep the blades engaged when running this compass-tuning exercise.

Here is a summary of the observations we just recorded:

Table showing the compass reading error observed on the 4 canonical headings. Notice that the logic added by the waynebaswell/ardupilot fork expects the compass adjustment in centi-degrees.

And now, assuming you’re running the waynebaswell/ardupilot fork (or you’re some future reader finding this article after this code has been merged into the main branch), you’re going to bring up Mission Planner –> Config/Tuning –> Full Parameter Tree –> AHRS and record the offsets we need (in centidegrees) in the AHRS_OFFSET parameters like this:

The waynebaswell/ardupilot fork gives you 4 additional parameters for fine-tuning the compass

Don’t forget to click “Write Params” to instruct Mission Planner to save the changes to your autopilot. Now if you take your robot back outside and point North, it should be true — or at least very close to true. Go ahead and do a sanity check on North-South-East-West with your robot.

If you’re following along on this build and you’d like to compare your ArduPilot parameters with the big Scag’s ArduPilot params, you can find the current parameter file attached at this end of this post.

Disabling the Internal Compass

The internal compass (in the Pixhawk 4 beneath the floor pan) is surrounded by way too much static steel and swinging blades to provide useful information. Here is what the Compass settings look like in Mission Planner:

Mission planner settings to only use external compass.

With these adjustments in place you should be able to get really solid heading readings out of a remarkably inexpensive mag.

The Finished Product

If you build robots you know that you’re never really finished — but the huge ArduPilot mower is performing quite well these days on increasingly long missions. Let’s wrap up the build description with a pic and a vid.

Here’s to Pops

You may have noticed that Pops made a slight cameo in the servo-danger-scenario pictures earlier — the truth is that most of the cool mechanical engineering on this linkage and, well, throughout this entire blog is largely either his thinking or the result of bouncing ideas off him. He used to run a big vessel fabrication shop, but now he has settled into teaching youngsters how to weld at the local High School. Not surprisingly, his students routinely take the lion’s share of trophies for craftsmanship quality at welding competitions.

Concluding Thoughts

Hopefully this little overview of an autonomous mower design helps out some future traveler.

If so, the words of Fred Brooks (in his great book) mirror your author’s thoughts:

Soli Deo gloria — To God alone be glory.

Your servant,

Roby

Footnotes

1. This is not at all a criticism of the efforts from those wise men almost 15 years ago — linear actuators were likely just fine for the controlled tests they were performing — besides, pulling off that conversion in the days of early-2000s internet is quite an achievement.
2. Grandma’s old adage that The Internet Doesn’t Forget apparently applies to everything except technical documentation and prior software/firmware versions.
3. If future readers end up needing to run these servos at 24v on an existing 12v system, just grab a 24v Step Up Converter with sufficient amperage rating.
4. Aspiring gardeners who were too busy in their youth swimming in Grandpa’s pond to help out with the tomato plants will find a wonderful primer in Charles Wilber’s nearly immaculate explanation of his methods. The late Wilber’s masterpiece is marred only by the publisher’s decision to contract out the book’s cover design to the guy who creates the packaging for all those As Seen on TV products. The book’s picture on Amazon looks goofy, but rest assured that the raving reviews of it’s content are legit and warranted. Modern readers will note that Wilber was several decades ahead of the popularity curve in his commitment to organic gardening (with a few exceptions) and treating the Earth with care.
5. Readers with understandable-but-hazardous superbike addictions may add several whole numbers to their probability of survival by giving at least a few afternoons to consider David Hough’s classic guide. To this day, thanks to Mr. Hough, when I’m driving down the road and see a car (hopefully) stopping / waiting at a stop-sign on a perpendicular road up ahead, I remember his counsel to look at the very top of that vehicle’s front wheel for an additional clue about whether that driver has seen you (being overlooked is the huge danger for motorcyclists). Hough’s reasoning is simple: when a vehicle at rest begins to move, human eyes will usually detect the rotation of the wheel before registering the relatively-small motion of the large vehicle’s body.

pixhawk4-scag-ardupilot-params-aug-22-2019.param

In ArduPilot world, the set of coordinates (or waypoints) that your rover is scheduled to visit is often called a mission (and this mission is stored in a plain text file called a waypoints file). If you inhabit that world, you know that the de facto standard tool for basically all robot configuration and mission building is the superb Mission Planner (Windows) application written and actively maintained for nearly a decade as an open source project by Michael Oborne, an Australian super-dev. (Incidentally, the Australian people seem to be making a disproportionately big contribution to the advancement of open source autonomous robotics.)

Mission Planner is a Swiss Army knife for ArduPilot-based robots — it provides dozens (if not hundreds) of functions — but for building the kind of centimeter-level side-by-side pass missions that I like to use for mowing, I don’t think it yet has functionality to fill that exact niche.

For a while I’ve wished for some browser-based / Google Maps-based / highly zoomable (i.e. for cm-level adjustments) app that will let you define a region via a standard Google Maps polygon and then magically generate a side-by-side pass (i.e. think parallel lines instead of concentric circles) mission for a mower to completely cover the user-defined area.

Not finding exactly this anywhere else, I’ve taken a rough stab at such an application and here present it to you for your enjoyment and critique:

Precision Mule: Serving up centimeter-level missions for your mowing rover

What is Precision Mule?

Precision Mule is a centimeter-level mowing robot mission path generator. I say “mowing robot” because the motivational use case for the routes it generates is a mowing autonomous rover.

You use Precision Mule like this:

Please note that I have only tested Precision Mule with Firefox and Chrome — if anyone would like to tweak the code to work on your fav browser, feel free to submit a PR on GitHub.

Examples

So, for example, here is a mission I built with Precision Mow to mow the west lot:

Here’s a time-lapse of the mowing rover we’ve been using lately running that mission yesterday:

Here’s a mission covering basically the same area, but instead of north-to-south passes, we’re doing east-to-west, i.e.:

Just as a reminder, the heading of the waypoints is determined by the line between the first and second points of your polygon.

Here’s a time-lapse from yesterday running that mission:

Here’s a mission covering the north part of my property:

And here’s a time-lapse from earlier this morning mowing that mission:

Look closely and you’ll see Cousin Jason dropped by to check out the run.

Precision Mule Deep Dive

If you’re still reading at this point, you must have a use for this software. Here’s a video I threw together to (hopefully) make your precision path generating experience as enjoyable as humanly possible:

It’s all Open Source

The backend java server: https://github.com/waynebaswell/precision-mule-server

The ui web client: https://github.com/waynebaswell/precision-mule-ui

Look forward to hearing your feedback.

Till we meet again,

Sincerely,

Roby

1. First-date-level excitement about the capabilities of the new component and the possibilities it will open up for your projects
2. Final-exam-level dread of the process of figuring out how to use the thing

#2 can be especially pronounced when you’re an early adopter and the internet hasn’t yet given birth to much documentation about your newly acquired component.

Since that fateful day in early 2018 when u-blox announced the ZED-F9P L1/L2/L5 GNSS receiver, I’ve logged an embarrassing count of visits to their website anticipating every new piece of information that they would publish about the little ZED.

Eventually u-blox announced they would call the official development board the C099-F9P. As an old-school believer in the technology maxim that you wait ’till the 3rd iteration of some new product before adopting, I had hoped that u-blox would quickly release a few updates to the dev kit so that I could feel like a wise old sage buying the more refined product.

Well, Digi-Key began selling the C099-F9P dev kit and I could only wait a month or so before caving in and buying two:

The familiar thin white rectangular box with the beautiful u-blox logo stirs long dormant Christmas morning feelings

Unboxing the C099-F9P High Precision GNSS RTK Development Kit

When you unbox your C099-F9P, here’s what you’ll find:

Historically, the component prices of RTK hardware have been astronomical — even the multi-frequency GNSS antenna (hidden in the nondescript brown box in the upper left) has been a thousand dollar USD component up ’till recently.

Hooking up the components yields this:

The ZED-F9P hooked up and ready for for action. Note the GNSS antenna is sitting on the circular silver magnetic ground-plane. The metal ground plane significantly increases antenna performance by reducing signal multipath issues. Nearly any metal ground plane (i.e. the roof of your car) will do the job — in general, the bigger the plane the better.

If you go on a late-night RTK GNSS wiki binge, you’ll likely have your brain numbed as you start to uncover the considerations involved in distilling the faint chirps from several dozen non-homogeneous satellites orbiting our planet 12,000 miles away into a centimeter accurate position in your front yard 20 times per second.

Let’s be clear: this blog post will not add to that scientific body of knowledge at all — instead, I’m assuming you’re a layman from some field other than RTK GNSS and you’ve stumbled across this ground-breaking technology as it’s effectively the only game in town for repeatable absolute centimeter-level outdoor positioning.

Those wishing to understand the technology underpinning RTK GNSS more deeply will find capable mentors in Tomoji Takasu^[1], Clive Turvey^[2], and Tim Everett^[3].

Shortcuts to High Precision GNSS with the Revolutionary ZED-F9P (or “Early Adopter Tax Evasion”)

This blog post should serve as a crash-course in getting your ZED-F9P rolling. Success on my part will be measured in helping you avoid much of the early adopter tax that I had to dutifully pay through the usual fog of frustration and confusion that eventually gave way to the thrill of using a game changing component.

At this point I’m going to suggest something that (I think) you’ll eventually thank me for: Go ahead and buy two C099-F9P dev kits so that you can stand up your own Base Station. Trust me, at some point this will almost certainly feel like the best $250 USD investment you’ve ever made.

We’re going to set up one of the C099-F9Ps as the “Base Station” (to provide “Corrections”) and the other as the “Rover”. You have the option to use some local correction source if available, or to use a subscription correction service (i.e. the dev kit comes with a free trial subscription to a remote correction service called HxGN SmartNet — I’ve not used). Your mileage may vary, but I’ve never regretted setting up my own base station and relaying the RTCM corrections on a dedicated radio link.

Let’s start with the simplest possible use-case — hard-wiring the corrections output directly from the “Base Station” board to the “Rover” board (via a male-to-male breadboard wire). Once you have the setup running with a wired RTCM connection, you’re just a pair of 3DR radios away from a long-range wireless RTK setup.

RTK success with two C099-F9P boards wired to each-other

Here are the high level items we’ll accomplish to get our self-contained RTK system running:

1. Connect to your ZED-F9P from within u-center
2. Update the ZED-F9P receiver firmware on both C099-F9P boards to the latest from u-blox
3. Set up one C099-F9P board as the Base Station, one C099-F9P board as Rover, and wire RTCM output from Base Station to Rover

1. Connect to your ZED-F9P from within u-center

On Windows 10 (no knowledge if this procedure works on other versions of Windows), before you plug in the C099-F9P, open up the Device Manager and expand the Ports item. Now plug in the C099-F9P and you should see 3 new ports that appear — one for the ODIN-W2 radio (COM7 in the example below), one for the ZED-F9P (COM6 below), and a USB Serial Device port that’s apparently some kind of all-in-one port where you don’t have to guess baud rates correctly. It’s much easier IMO to use this port (COM3 below).

On Windows 10, I seem to have the best success connecting to the “USB Serial Device” port that appears when you plug in the C099-F9P

From within u-center (I’m using u-center 19.04 which is the latest as of May 2019), click the down arrow next to the port icon in the upper left and you should see a port number corresponding to the USB Serial Device you noted above (COM3 for me). Select this port and we’re now connected to the dev board.

2. ZED-F9P firmware update

Updating to the latest firmware may not be required, but it’s quick and easy — so I’d say go for it.

Here’s a little video I made of updating the ZED-F9P firmware.

3. Set up one C099-F9P as the Base Station, the other C099-F9P as the Rover, and wire RTCM output from Base Station to Rover

You’ll need to know your Base Station antenna’s coordinates to complete this section. If you’re not familiar with how to get those coordinates, check out Addendum 1 below.

The configuration procedure we follow in the video below configures the receiver to output NMEA PVT (Position, Velocity, Time) to the UART2 (June 2019 update: output on UART2 has been spotty — so I’m just using UART1 for PVT output) UART1 Tx port at 5Hz (i.e. a new reading every 0.2 seconds) at 115k baud. It also configures the UART2 Rx port to receive RTCM corrections at 115k baud. These settings are convenient for the mower rover we blogged about last time. Note that you can set the output rate up to 20Hz, but you may need to bump up the baud to something higher than 115k (otherwise the line may get clogged up and the readings would be delayed). On the slow rover we’re using to cut the grass, 5Hz is plenty and the Pixhawk 4 we’re using communicates nicely at 115k baud.

If you’ve followed the steps above, then you can easily send RTCM correction data from the Base Station to the Rover by connecting the the Base Station UART1 Tx to the Rover UART2 Rx with a simple breadboard wire.

Caveat: the one-wire approach will work if you’re powering both boards with an electricity source with a common ground. If not, you can just connect the two boards’ GND sockets together (you’ll find multiple GND sockets on the dev board) with another breadboard wire.

It wasn’t obvious to me from the documentation which sockets on the C099 board map to UART1 and UART2 — perhaps all other humans were born with this knowledge. Here’s a little mapping diagram that will hopefully help future travelers:

Showing the u-center screens for the ports on the Base and Rover. Note the curious black lettering (like “J3 >1” and “J9 > 2”) — the official C099-F9P documentation adopts the names “J2”, “J3”, “J8” and “J9” for the terminal blocks — just showing what the terminal is called in the C099-F9P u-blox documentation.

You’ll note that we’ve taped off the ODIN radio antenna jack (remember that I don’t use the on-board ODIN radio). This just gives me one less thing to remember when connecting the GNSS antenna to the dev board.

ZED-F9P real world performance demonstration

Here’s a little video showing why this small RTK receiver is so significant:

Rolling our own long-range corrections link

If you’ve been using u-blox high precision gnss components since the NEO-M8P, then you may have previously used the superbly easy-to-use and capable C94-M8P development kit for that (L1-only) module^[4]. The C94-M8P basically did everything right from an ease-of-use standpoint for a builder wishing to probe that gnss module’s functionality. For instance, the development kit was sold as a pair of modules (i.e. so that you weren’t tempted to skip setting up your own local base station), the radio was based on SiK telemetry achieving great range, and the procedure for setting up a base station / rover system was very simple and well documented^[5].

I spent several unrecoverable hours trying to get two C099-F9Ps to communicate with each-other via the on-board ODIN-W260 radio. Efforts tried include downloading multiple versions of the s-center radio evaluation software, feverishly googling for any kind of documented success story setting up the link, setting jumpers, and flashing various firmwares and configuration settings to the dev board. I’m not sure if you’d describe my efforts as total failure or absolute failure, but I never saw any sign that I was close to actually making the communication link work.

Fortunately, long-range radios have become very cheap in recent years, so snag a pair off eBay, Amazon or SparkFun and you’ll have a long-range RTCM corrections link between the Base Station and Rover with minimal effort.

3DR radio configuration

Mission Planner has built-in functionality to configure the radios.

2 SiK radios connected to the Mac (running Windows via VM). Note that some SiK radios don’t have micro-usb ports — you connect them to your PC’s USB via a USB-to-Serial adapter.

Here’s a close-up of the relevant screen in Mission Planner, along with the settings of the radios that I’m using for the link:

Now that the radios are communicating, it’s time to remove the RTCM wire connecting the ZED-F9Ps to each-other and instead connect them wirelessly via the radios.

C099-F9P dev boards powering 3DR radios for wireless long-range RTCM corrections.

That’s all there is to it! Note the solid yellow light on the Rover indicates it has an RTK fix — I love seeing that light.

Sincerely yours,

Roby

Addendum 1: Finding Latitude/Longitude/Elevation for your local Base Station Antenna

This simple fact, combined with both a rapid fall of hardware component prices and a steady refining of history altering open-source autopilot stacks, means that many passerby have begun seeing sights previously uncommon to man.

Exhibit A:

In this blog post we’ll introduce the main components of our mowing rig. In future posts — hopefully not 2 years from now — we’ll try to flesh out some more system details. As always, feel free to post questions in the comments and I’ll try to address anything I haven’t explained well.

The Pursuit of the Perfect RTK GNSS System

As we’ve previously discussed, for several years the long $$ poll in the hardware tent hamstringing our autonomous mower dreams has been centimeter-level robust GNSS gear¹.

2 years ago we marveled that ComNav had built a precision GNSS module for ~$1000 USD capable of positioning to within 1cm at 10hz (i.e. 10 readings per second — or a new position output every 0.1 seconds). Our excitement was justified because the big agriculture/surveying GNSS vendors were concurrently exacting perhaps 5-10x that price for similar technology.

It is with considerable excitement we report that now in 2019, your affordable autonomous field-navigation robot dreams are more approachable than ever.

The reason for this, of course, is largely the gift that u-blox has recently introduced to mankind: the ZED-F9P L1/L2/L5 GNSS module.

Example hardware components for an Ardupilot based precision auto mower

If you live thousands of miles away from Switzerland like me, then you may also shamefully admit that your knowledge of that beautiful country has hitherto been effectively limited to the following:

1. The Alps
2. Remarkable banking
3. Swiss Army knives

Well sister, get ready to add a 4th item to that list:

Who are u-blox and why do they matter?

u-blox is a Swiss publicly traded, highly profitable, growing, financially healthy, dividend paying fabless chip designer that is quietly introducing a line of low-cost ultra high precision GNSS modules that will likely change the entire world².

Over the past 20 years, u-blox has been developing an ever improving line of high performance GNSS receivers that have become the de facto standard GNSS component in several important markets.

If you’re not very familiar with the world of drones here’s a way to think the place u-blox occupies in that market:

u-blox is the Intel of IoT GNSS

u-blox really is that dominant as the provider for drone GNSS receivers, and it’s reasonable to speculate that they’ll soon provide the high precision GNSS modules onboard a big percentage of the ~70 million automobiles annually sold worldwide.

Many more glowing words could be accurately penned about u-blox, but our task at hand involves using their ZED-F9P high precision GNSS module as the position solution for an autonomous mower.

So what’s the big deal with the ZED-F9P?

In laymen’s terms, the ZED-F9P allows any outdoor thing to know it’s exact position on the face of the earth to within ~1cm with up to 20 position updates per second at a price-point that’s an order of magnitude lower than similarly robust incumbent solutions.

It does this for a flat one-time price of ~$500 USD. That price is for 2 ZED-F9P modules (along with 2 antennas) to build a stand-alone system (i.e. one of the ZED-F9P modules is set up as a Base Station / “correction” provider). If you have an existing correction source, then you’ll only need to purchase one module, and so your cost is halved to ~$250 USD.

The fine citizens of Precision Ag / Land Surveying world, arriving at this page and reading the ZED-F9P specs and pricing, will likely have a visceral appreciation for the u-blox accomplishment that perhaps few others can fully understand⁷.

For $500 you can purchase a 2 receiver / 2 antenna (a capable active patch antenna is included in the dev kit) L1/L2/L5 multi-constellation 20hz system that’s comparable in performance (and arguably much more approachable in terms of ease-of-integration) to a $10,000 system from one of the existing old school RTK manufacturers.

Readers familiar with RTK technology jargon may be interested in a flyover of the ZED-F9P’s technical specs:

ZED-F9P Technical Highlights

Price: ~$250 USD³

Multi-Constellation: GPS / GLONASS / Galileo / BeiDou — In other words, it’s processing position signals from satellites from all 4 of the major GNSS constellations.

Multi-Frequency: L2OF, L2C, E1B/C, B2I, E5b, L1C/A, L1OF, B1I — This is the real kicker — historically you have only obtained multi-frequency receivers by parting with thousands (or ten-thousands) of USD dollars.

20HZ output capable: Stated differently, just to be clear, it can give you 20 unique cm-level accurate positions every second.

Well-documented and mature presence: i.e. autonomous autopilot stacks (such as the gold-standard Ardupilot) have supported u-blox gnss receivers for years⁴.

Supports standard RTCM corrections.

Low power consumption.

If you’re coming from the world of Precision Ag or Land Surveying and you read those specs, due to your experience with the companies that have supplied that market for 2 decades, you may be thinking “yeah but how much does The Man charge me for software un-locks for all 4 constellations, multiple frequencies, and 20HZ output?”

Answer: 0.00 USD.

Yes, the ZED-F9P comes with all those features out of the box — the only catch I know of is our shared inability to stockpile these modules and then time-travel back to 2015, making a fortune through resale.

Rover Build

The rover we’re describing today is built on a repurposed electric wheelchair powertrain as we’ve previously detailed. But note that we’ve upgraded a few components since the previous discussion about this rover.

The significant modifications are these:

1. Replace our beloved ComNav K501g with the superior u-blox ZED-F9P.
2. Upgrade rover wifi system with Ubiquiti mesh network⁵ (i.e. we run a Ubiquiti access point on the the rover that’s joined to our local Ubiquiti network. This offers many benefits, for example: it’s now trivial to shell/VNC into rover’s onboard Raspberry Pi from any computer on our local network).
3. Replace Pixhawk 1 autopilot with Pixhawk 4 autopilot. Just to be clear, this is not because the Pixhawk 1’s functionality is yet significantly outdated for this application (even though Pixhawk has been around since 2013) — instead, when Holybro introduced the Pixhawk 4 I got curious about it’s performance, and liking it’s specs (and Holybro’s solid reputation) but not finding many user reviews of the module, curiosity overtook and I just bought the little guy from Sparkfun.

Rover front view from above

Note the next-level red/black clamps connecting the + – battery terminals to the AC/DC inverter

The box on the left is the power box — box on the right is the brains box.

The “RTK” radio at the bottom is dedicated to receiving RTK corrections from the base station.

Another angle of the brains box.

Mowing Deck Build

We hacked together a mowing rig⁶ as follows:

1. After eyeing the big 60″ deck on our old Scag mower, we cut out a similarly-shaped (albeit scaled-down) hexagon from 1/8″ sheet metal.
2. Procure 3 24v 150W ZY6812 150-GM150115 DC motors from eBay (somehow those powerful motors are shipped to your door @ $25 a pop).
3. Hook up some shaft coupler extensions (from eBay) to the DC motors. This buys us a couple extra inches of clearance between the mower deck and the blade disk. Note that I had to Loctite the nuts that snug these extensions to the motors. Also note that if you want to purchase extra nuts for this extension (for example, if you want to hold those Honda disks in place by sandwiching the blade disk between two nuts), that the nut size ia a little odd: M10 Left Hand 1.5. This Amazon query should get you what you need.
4. Bolt the 3 DC motors to mowing deck sheet metal.
5. Attach Honda blade disks to the shaft coupler extensions. I like the Honda disks because they’re metal (and because they’re Honda). I had to drill out the hole in the center of the disk slightly — i.e. you have to make the hole’s diameter a little bigger to use the shaft coupler extensions that we’re using.
6. Ask Pops to weld together a carriage structure to attach to the rover. If you study carefully the metal and swiveling wheels on this structure you may recognize they’re just parts we’ve salvaged that were laying round from prior power-chairs.
7. Suspend the mowing deck from the carriage by 4 loops of 1/8″ galvanized 7×19 steel wire cable (this ultra-strong wire rope was leftover from building the backyard containment fence a few years back). Those aluminum brackets attached to the deck that the cables loop through were laying around the shop.
8. Run ground wire and fused 24v power wires back to the mowing deck.
9. Create mower deck sides by bending & cutting sections of 3″ Aluminum flat bar. Bolt aluminum sides to deck using these brackets.
10. Bolt two Spiderman bicycle training wheels to front left/right sides of mowing deck — this helps the deck to “float” along without getting hung up on uneven terrain.

Honda blade disk part #72612-VP7-C50

Blades (not pictured) swivel on 3 bolts welded to the disk

Wrapping Up

Let’s bring this post to a close with two time-lapse videos (one run — but video taken from 2 cameras — note that the second camera’s battery died before run completion) taken this morning.

Note the number of obstructions that would challenge an RTK system of lesser capability.

The mowing rig works quite well, keeping the back yard neatly trimmed and letting my clothes stay clean in the process.

Sincerely yours,

Roby

June 27, 2019 update — Here is my current Ardupilot Rover Pixhawk 4 param file for the rover: Pixhawk4-Rover-Params

In that post I couldn’t quit raving about the ComNav K501G card. The reason for this, of course, is that the card is awesome.

Not awesome is the temporary dearth of information explaining how to use the K501G.  I say temporary, because right now we will commence building the internet’s knowledge base for working with this superb card.

The Case of the Missing Quick-Start Guide

Back in 2016 when I was on the uncharted frontiers of K501G exploration, ComNav published a quick-start guide that proved quite helpful. For some reason they no longer offer it on their site, but I found it in my archives, and am posting it here in hopes that it helps you out as well.

K501G Quick Tour

While we’re at it, let’s go ahead and post an archive link to the K501G v1.5 reference manual:

ComNav OEM Board Reference Manual_v1.5

And while we’re still at it, here is a link to ComNav’s canonical documents.

Using the “Dev Kit”

Dev kit sans antenna

If you’ve found yourself in possession of a “dev kit” for the K501G, then you likely have a setup like the one pictured above. I forgot to include the GPS antenna cable and GPS antenna in the above picture. The GPS antenna, of course, connects to the TNC Female barrel connector on the dev board above. The dev kit is simply there to make it easier to configure and test the K501G. If you’re on a really tight budget, we’ll show below how you can get by just fine without a dev kit.

The big idea to keep in mind when working with the K501G board (other than don’t kill it with static), is that you can communicate with the board on any of it’s three I/O ports. These ports, unsurprisingly, are called COM1, COM2, COM3. Later on we will give the K501G instructions for what kind of output to write or input to accept on each of these ports.

You’ll notice in the dev kit above that the dev board uses three 9 pin RS232 connectors for connecting to COM1/COM2/COM3. I have no clue why the RS232 connectors are used (instead of, say, a straight USB connection) — perhaps it’s a joke, maybe they wanted you to think about the 90s and get nostalgic — more likely, I suppose, there’s some legacy reason why dev boards were made this way years ago and it’s now easier to ship RS232-to-USB adapters rather than design a new dev board.

Mounting K501G to dev board

Go ahead and mount your K501G to the dev board, connect the 9 pin RS232-to-USB connectors to your dev board, connect the antenna to the dev board via a TNC-Male-to-TNC-Male antenna cable, and connect the power supply to your dev board. You’re now ready to flip the “On” switch on the dev board.

In the dev kit setup pictured above, we connect the 9 pin RS232 connectors to COM1 and COM2. You may be wondering why the board has 3 communication ports. I can’t speak for the manufacturer’s motives, but I can say that having 3 ports allows you to cleanly segregate the duties of each port.

For example, take a look at the writing on the blue tape on one of my K501G boards:

K501G with port function and baud rate labeled on blue tape

Notice on the blue tape above I’ve labeled the function of the ports as follows:

1. Port 1: PVT (Position, Velocity, Time) OUTPUT
2. Port 2: RTCM Correction INPUT
3. Port 3: Configuration (I connect to this port to configure the board — in theory you could connect to port 1, but with 10Hz RTK output, the port will likely become congested).

If you’re going to spend much time with these boards, you may find that labelling the port’s function spares considerable future confusion.

Configuring the K501G with Compass Receiver Utility (CRU)

In order to configure the board, we’re going to use the software ComNav provides called either “Compass Receiver Utility” or “CRU OEM Board Control Software” (direct download link, download page link). Compass Receiver Utility (hereafter called CRU) is not particularly intuitive for a NON-GNSS-Professional user. I’ll attempt to give you the high points you’ll need so that you can get your board configured as either a rover accepting corrections and outputting PVT data or a base station outputting RTCM3 corrections.

For the following instructions, I’m assuming you’ve connected your Windows PC to PORT 3 via the RS232-to-USB adapter. Why port 3, you ask? No technical reason, just as a habit I like to always configure the board via port 3. We will configure port 1 and port 2 like I configure my boards (remember the port descriptions above the beautiful picture above).

Be sure that you’ve turned the dev board’s power button on – maybe it’s just me, but I’ve somehow overlooked the power button on the dev board more often than I’d like to admit.

Once you’ve downloaded and installed CRU, you’ll want to open it up.

CRU “Set Port” Screen

In the upper left hand corner of CRU, click the “Set Port” turquoise button. In the “Connection Settings” dialog box, select your COM port from the “COM” dropdown box. Next select the appropriate baud rate from the “Baud Rate” dropdown box. I believe the default baud rate is 115200 for the K501G. Let’s assume that baud rate is correct, now click “OK”.

It turns out that CRU is quite ambitious and once you click “Ok” it establishes a connection to the COM port you selected (no need to click the green “Connect” button in the upper-left next to the “Set Port” button).

Now let’s verify that you’re actually connected to your K501G board.

In the CRU, click the black “Command” button near the top/center. This will bring up the “Command” dialog window pictured below. You can type commands into this window and click “Send” to send them to the K501G board. Here’s the thing:

YOU MUST CLICK “ENTER” ON YOUR KEYBOARD BEFORE CLICKING THE “SEND” BUTTON.

YOU MUST CLICK “ENTER” ON YOUR KEYBOARD BEFORE CLICKING THE “SEND” BUTTON.

YOU MUST CLICK “ENTER” ON YOUR KEYBOARD BEFORE CLICKING THE “SEND” BUTTON.

If someone had just pointed that out to me (and if I had remembered it) it would have saved me hours of frustration. In programmer speak, the board does not process a command until it sees a \n (newline) character.

Let’s begin by showing the WRONG WAY to verify your connection:

FAIL: log version command sent without hitting the ENTER key on your keyboard before clicking the “Send” button

Notice in the image above that the “Terminal” tab at the top displays both the command you send to the K501G (via the Command window) and the response from the board.

Let me push on this a little more so you’ll remember it. If you type in log version and forget to append the newline character (by pressing ENTER on your keyboard), you will see the following in the terminal tab:

FAIL: Tell-tale sign is the </command> closing tag on the same line as the command

You are not going to forget to press ENTER. Therefore you, my smart friend, will be seeing this screen:

SUCCESS: log version — if screenshots captured cursors blinking, you would see the cursor blinking on the line BELOW log version

Notice the output in the Terminal tab above. If you see output that looks like this “<VERSION COM3 0 60.0 FINESTEERING…” then you are in business.

It’s quite frustrating that the CRU doesn’t append the newline character for you, but hey, you wouldn’t want it to be easy would you? I think half the battle of working with the K501G card through the CRU is simply specifying the correct port/baud rate and then remembering to append a newline character to your commands.

K501G RTK Rover Commands

Let’s first discuss the configuration necessary to instruct your K501G to output RTK position at 10Hz on port 1 and to accept RTCM corrections on port 2. Cutting to the chase, here are the commands:

FIX NONE
REFAUTOSETUP OFF
UNLOGALL
SET CPUFREQ 624
SET PVTFREQ 10
SET RTKFREQ 10
LOG COM1 BESTPOSB ONTIME 0.1 0 NOHOLD
LOG COM1 BESTVELB ONTIME 0.1 0 NOHOLD
LOG COM1 PSRDOPB ONTIME 1
INTERFACEMODE COM2 AUTO AUTO ON
SAVECONFIG

Let’s give a little color to those commands before moving on.

FIX NONE — If the rover was previously set up as a base station, this clears that out. I once lost several hours trying to figure out why the K501G wouldn’t record movement for more than a minute after startup — turns out it was previously configured as a base station. FIX NONE simply tells the K501G to stop running around pretending to be a base station.

REFAUTOSETUP OFF — Tell the board to not configure itself as a base station. I have no idea if this is necessary, but after the incident above, I became obsessive about ensuring that a rover card isn’t configured as a base station.

UNLOGALL — Instruct the K501G to stop all output on all ports. We just enter this enter this command to ensure we’re dealing with a clean slate — we’ll specify all output below.

SET CPUFREQ 624 — Bump up the K501G’s CPU frequency to handle 10Hz RTK computation.

SET PVTFREQ 10 — Compute PVT data at 10Hz.

SET RTKFREQ 10 — Compute the RTK solution at 10Hz. You may find it odd that you have to specify both PVTFREQ and RTKFREQ. Why is this necessary, you ask? You know better than ask your Mother and I these kinds of questions. Note: ComNav’s documentation says that RTKFREQ should not exceed PVTFREQ.

LOG COM1 BESTPOSB ONTIME 0.1 0 NOHOLD — Ouput the RTK position data to port COM1 every 0.1 seconds. In other words, output RTK data @ 10Hz. If you find the nomenclature of referring to port output as “LOG” a little confusing, then you’re not alone.

LOG COM1 BESTVELB ONTIME 0.1 0 NOHOLD — Output the velocity data to port COM1 every 0.1 seconds.

LOG COM1 PSRDOPB ONTIME 1 — If you want to see the DOP (Dilution of Precision) data every second, then this command is for you.

INTERFACEMODE COM2 AUTO AUTO ON — Instruct the COM2 port to accept RTCM/RTCM3 correction input.

SAVECONFIG — IMPORTANT!!! You have to enter SAVECONFIG for the board to save your changes permanently to non-volatile memory. Stated another way, if you don’t enter SAVECONFIG, the commands you enter will only be in effect ’till you restart the board — at which point the board will revert to the previously saved configuration. Depending on your perspective, I suppose, you will either regard this as a burdensome extra step, or as a slick feature that leaves the door open for bailing out if you’re ever unsure that you’ve entered invalid commands. At any point before you enter SAVECONIG you can simply cut power to the board and all your commands are quietly forgotten.

FINAL REMINDER: Suppose you copy the commands above into the CRU command window, but forget to click ENTER on your keyboard, and thus fail to append a newline character after the final SAVECONFIG command. Here’s what will happen: the K501G will process the first 10 commands (“FIX NONE” through “INTERFACEMODE…”), then it will not process the SAVECONFIG command. The reason for this behavior, of course, is that the first 10 commands have a trailing newline character, but the final SAVECONFIG command didn’t have the trailing newline character.

Now spend a few more seconds working out another really important implication of this: since the board didn’t process that final SAVECONFIG command, as soon as you restart the board, it will revert to your previously saved configuration — discarding all the changes you thought you had made.

K501G Base Station Commands

Now suppose that you want to set up your own base station so that you’re not relying on any external system for corrections (and presumably so that your baseline to the correction station will be shorter). We will consider two different configurations, first we’ll consider the scenario where you want the base station to automatically determine it’s antenna’s location every time it boots up (absolute accuracy to maybe 1-2 meters), next we’ll consider the scenario where you want to explicitly configure the base station’s antenna’s position (absolute accuracy basically as good as your surveyed point).

Both configuration sets presented below output corrections at 1Hz for both the GPS and GLONASS constellations in the RTCM3 format.

July 2019 u-blox ZED-F9P compatibility update

If you’re wanting to use your old K501G cards as base stations for a ZED-F9P rover, here are the messages you’ll want to log (i.e. don’t worry with logging any other messages we mention down in “Scenario 1” or “Scenario 2”):

LOG COM1 RTCM1005B ONTIME 1
LOG COM1 RTCM1074B ONTIME 1
LOG COM1 RTCM1084B ONTIME 1
LOG COM1 RTCM1230B ONTIME 1

End July 2019 update

Scenario 1: Auto Determine Antenna Location

Absolute accuracy: +- 2 meters

Relative Accuracy (accuracy of rover relative to base station): +- 1 centimeter (add 1mm per km distance between rover and base station)

Here are the commands to set up a base station that automatically determines it’s present location every time it boots up:

UNLOGALL
FIX AUTO
LOG COM1 RTCM1004B ONTIME 1
LOG COM1 RTCM1006B ONTIME 1
LOG COM1 RTCM1012B ONTIME 1
LOG COM1 RTCM1008B ONTIME 5
LOG COM1 RTCM1033B ONTIME 10
SAVECONFIG

Scenario 2: Explicitly Declare Antenna Location

Absolute accuracy: +-1 centimeter [depending on 1. Accuracy of surveyed point and 2. Distance of rover from base station (add 1mm for every km between rover and base station)]

Relative Accuracy (accuracy of rover relative to base station): +- 1 centimeter (add 1mm per km distance between rover and base station)

Here are the commands to set up a base station with explicitly declared coordinates:

UNLOGALL
FIX POSITION 28.9823853 -84.2492042 43.89
LOG COM1 RTCM1004B ONTIME 1
LOG COM1 RTCM1006B ONTIME 1
LOG COM1 RTCM1012B ONTIME 1
LOG COM1 RTCM1008B ONTIME 5
LOG COM1 RTCM1033B ONTIME 10
SAVECONFIG

You know that you’ll want to replace the 28.9###### -84.2###### 43.89 with your own latitude longitude height-in-meters values.

That wasn’t too bad, was it? I think a line-by-line explanation of the above commands would be overkill — we’re just telling the card to write the different RTCM messages to COM1. If you’d like an overview of the different RTCM messages, here’s a nice link.

Putting on Your Surveyor’s Hat

You may be wondering how you can get an accurate position of your antenna. Here are a few options:

1. Pay to have a location surveyed.
2. Make a quick-n-dirty guess by zooming in to the location on a site like this.
3. Generate your own super accurate near-survey-quality position with the CRU software via a publicly available corrections source and the K501 RTK card you configured above.

Let’s talk about how to make #3 happen. First you’ll need to find a correction source that’s within, say, 50km — the closer the better. Here is a list of public corrections sources. Here is a big beautiful map of public corrections sources. So, for example, here is the list of public stations for Alabama. If you look at the Alabama link, you’ll notice they give you one IP (205.172.52.26) and then a list of ports to choose from for various location / correction format combinations. I live near Foley Alabama, and I would like to receive corrections for both the GPS and GLONASS constellations in the RTCM3 format. Scrolling down the Alabama RTK pdf, you’ll notice that port 19405 is exactly what we’re looking for.

So now that we know the PORT and the IP for our corrections, it’s time to open up the CRU software and feed those corrections into our K501G RTK card. Remember that when we configured the RTK card above, we specified that we wanted position output on port COM1 and we specified that we would feed in corrections on port COM2.

Here we have another headache with the CRU software — AFAIK you can only connect to 1 port at a time. What this means is that you’ll have to open 2 instances of the CRU — one to read the PVT data from port COM1 and another to feed in the correction data to port COM2. So go ahead and open up two instances of CRU.

Ensuring that the dev board is plugged in and turned on, connect the (USB-to-RS232 adapter / 9Pin cable) from your PC’s USB to port COM2 on the dev board. Now open one of your CRU windows and connect to the dev board by clicking the turquoise “Set Port” button in the upper left.

Assuming you’ve connected to port 2 successfully blue “Diff” icon near the top left-center of the CRU screen.

Diff dialog box before connect

When we click click “Diff” we see a dialog windows called “Diff Source Setup”. Here we will enter the Host IP (you may be able to use a DNS name here like “myslickcorrectionsource.nasa.gov” — I haven’t tested it for DNS) and Port from above. Let’s cross our fingers and click the “Connect” button.  If all goes well, here is what you’ll see:

Successfully connected to correction source

Notice at the bottom right of that screen you’ll see some green text that says “Diff 363 B/S“.  You’ve likely guessed that means that you’re receiving corrections from the source specified at 363 bytes per second. That feeling you feel is the warm feeling of free corrections flowing in. That other feeling you feel is anger at having to open up a second CRU window. OK, let’s get over it and open up the second window and connect to port COM1 which we previously configured to output PVT data at 10Hz.

Here’s a screenshot of both windows open in parallel:

Two CRUs open — left screen is feeding corrections into port COM2, the right screen is receiving PVT from port COM1

On the CRU screens above I made a few clarifications in red letters. On the CRU screen on the right, when you see the “Position” field’s value reading “NARROW_INT”, that’s K501G speak for RTK Fix. You will notice several hundred points charted at the bottom of that screen are all falling within a 1.5cm radius. Not too bad when you consider we’re on a 16.29km baseline in the middle of the day, on an antenna mounted ~1m above the ground, flanked by tall trees to the north, powerlines to the north, a power transformer to the northeast, powerlines to the east, a big brick house to the south, and various other obstructions as pictured below:

Listen my friends, I’ve heard from several manufacturers since writing the original RTK article, BUT I HAVEN’T HAD ONE OFFER FROM A MANUFACTURER OF EITHER A GPS-ONLY OR L1-ONLY CARD WILLING TO COMPETE WITH THE K501G. I’ll let you guess why that is.

If you think the above pictures represent difficult conditions for a card to hold a RTK fix, then you’re not very familiar with modern L1/L2 GPS/GLONASS systems. I will keep repeating the same line: go ask the surveyors if you don’t believe ‘ole Dad Roby. If you don’t believe them, just buy the cards yourself and brace for awesome.

Speaking of surveyors, and getting back to establishing your antenna’s absolute location, if an actual surveyor were present, they would record several minutes of observation data at your antenna’s location. Then they’d process the data and would likely shave a few millimeters off the error. If, however, you’re OK with knowing your base station’s absolute position on the face of the earth to within a centimeter or two, then the above method should serve you well.

Once the CRU reports “NARROW_INT” (RTK Fix), just note the Latitude, Longitude, and Height displayed and you’ve got yourself an absolute cm-accurate antenna location.

While we’re pointing out quirks with using the CRU software, let’s point out another. If you’re using the CRU to observe data from a K501G at 10Hz (like the right side CRU instance in the dual CRU desktop picture above), within a minute or two the CRU will become bogged down and virtually unresponsive. I suppose the CRU falls victim to a good ‘ole memory leak. A hack to keep the CRU from freezing if you’re wanting to observe PVT data for more than a few minutes is just to dial back the frequency of the output, like this:

LOG COM1 BESTPOSB ONTIME 1 0 NOHOLD
LOG COM1 BESTVELB ONTIME 1 0 NOHOLD

That scales back the position and velocity output to 1Hz. Keep in mind that the card only saves changes when you send the SAVECONFIG command — therefore, you can enter the commands above and play around with CRU all you want, and when you’re done, cycle the power and it will go back to outputting RTK data at 10Hz.

High Risk Move: Ditching the Dev Kit

Earlier I told you that I’d show you how to ditch the dev kit and perhaps save a few bucks. Since the ComNav is just outputting serial data, you can communicate with it via a USB-to-TTL adapter. To connect your antenna to the board, you’ll want a TNC Female to MCX Male adapter like this. The K501G’s pins are 2.0mm pins and they couple with a 2x10p 2.0mm Dupont Connector. If you don’t have a 2.0mm crimping tool laying around, you could just pick up some 2.0mm to 2.54 mm Dupont wires to use for hooking up the K501G to your USB-to-TTL adapter.

Here is a hacked-together base station following this kind of approach:

Note in the setup above we’re only listening to COM1 on the K501G (we’re receiving corrections on the green wire). The red and orange wires supply power to the board and to the antenna. The black wire is ground. If we wanted to talk to the board, we’d hook up the blue wire to our USB-to-TTL adapter. The function of each pin is covered in the K501G Board Specification.

You may notice in the picture above that I label “1”, “2, “19”, “20” corresponding to the position of the respective numbered pins. I’m paranoid that I will get pin 1 and pin 20 mixed up (and hook up the wrong wires to the pins) and so I write all over the boards to try to mitigate that risk. As always, the labels seem to have a better memory than me.

Of course, you’ll want to be very careful handling and wiring up the board. If you purchased the K501G with your own hard-won cash, then that reminder is unnecessary.

Wrapping Up

After last month’s RTK post, I felt like I owed it to any readers who would venture to purchase the K501G to give you a guide that would, perhaps, save you a fair bit of time and headache when setting up this card. Configuring the K501G is the least fun part of owning it. Fortunately, if you get it right, you shouldn’t have to do it again for a long time (maybe forever).

While I feel like I’ve now discharged an obligation, it may be that this guide leaves you with more questions than answers. If anything is unclear, just leave a question in the comments and I’ll try to respond and/or clean up any ambiguous parts of the article.

I don’t know about you, but I’m ready to get back to building robots.

Until next time,

Sincerely,

Roby

Grandaddy was born on May 20, 1937 in rural northwest Alabama. He grew up in a poor time in a poor state in a family that was poor relative to other people. He spent his childhood working on a farm, plowing behind mules, bent over picking cotton, waking up early and staying up late taking care of pigs & chickens & whatever else was necessary to provide the next day’s food.

At school, during lunch, the food Grandaddy brought reflected his family’s financial status. The more affluent people could afford to purchase sliced bread, but those at the lower end of the economic spectrum would bring homemade bread or biscuits. Grandaddy brought homemade bread — perhaps there’s some connection that, years later, I remember he started making homemade bread as a hobby. He gave so many loaves of sourdough bread away to anyone who take it, never accepting payment, just happy to give and serve.

At the age of 18, he married my Granny, and over the next 10 years they welcomed the birth of 3 daughters. Early in his 20s Grandaddy decided to devote his life’s work to preaching. Within a few years he moved with Granny and his 3 young daughters to the church in Robertsdale where they settled down and where he has now served without interruption for over 50 years.

It’s perhaps ironic that I’m writing this tribute to my grandfather on a blog about robotics. Grandaddy has no interest in technology, he does not have internet access, has never purchased a home computer and he does not carry a cell phone. Readers of this blog would likely experience withdrawal symptoms if they chose to forego any items on the preceding list. But Grandaddy has served people for all these years without those electronic staples of your life and mine — and I’ve never known of anyone who has lived more fully.

While I can’t imagine a day disconnected from the internet, Grandaddy has rarely lived a day disconnected from lives of those who need someone to talk to, need someone to see, need someone who is present. He’s logged over 250,000 miles on his favorite car — a 2003 Toyota Camry — so many of those miles on seemingly unending trips to the local nursing home, to the local hospitals, to the church and to the cemetery.

I think a story will give you an idea of the kind of man my Grandaddy is. Recently a friend of my dad was diagnosed with cancer. Grandaddy did not know this man very well — he does not live in our town and he is not a member of the church where Grandaddy serves. But when Grandaddy heard of the diagnosis, he jumped in his car and drove 45 minutes across Mobile Bay to the hospital, just to spend a few minutes with the man and with his family. I only know of this story because the man’s wife called my mother and told her that Grandaddy had just visited them in the hospital — she was profoundly moved by how kind Grandaddy was, and how considerate he was to reach out to them in such a difficult moment.

If you were to meet Grandaddy, you might be surprised to find that he is genuinely interested in what you have to say, he makes eye contact, he keeps a kind and understanding countenance, and he listens. If Grandaddy were to find out that you were going through a difficult time, he will show up or he will call — whichever he thinks you would prefer. He and Granny will offer help, they’ll listen to you and pray for you. No strings attached, no sales pitch, no invasive questions. Just care. Just concern. Just love.

In a time when many of us can accurately chart the Kardashian family tree, but have no idea of our neighbors’ names, Grandaddy stands as a reminder that perhaps we need to reconsider the trajectory of our lives. I doubt Grandaddy has ever heard of the Kardashians, but he knows all of his neighbors’ names. He knows their children’s names. Over the years, as many of them have died, nearly all of their families call on Grandaddy to give the eulogy at their funeral.

Those who hold positions of leadership in a town, or a business, or a church are often unable to please everyone. But I’ve never known of anyone who was able to successfully hold a long-term grudge against Grandaddy. Years ago, Grandaddy unknowingly moved the corner post demarcating the property line between his land and his neighbor’s. Over the next several months he noticed that this neighbor would no longer wave or talk with him. Grandaddy was baffled by the neighbor’s sudden change of behavior and was puzzled by what might have happened. Finally Grandaddy asked the neighbor if something was wrong. The neighbor asked Grandaddy why he had moved the corner post — Grandaddy had no idea that the post he had moved was the property corner post. He profusely apologized and immediately asked the neighbor to tell him where the corner post should be reset, and Grandaddy had the post placed exactly where the neighbor wished. No survey needed — Grandaddy was happy for the neighbor to decide the property line if that meant winning back a friend. From that point forward the neighbor and Grandaddy were back on great terms. Years later, of course, the neighbor’s family called on Grandaddy to preach his funeral.

Grandaddy can offer a gentle rebuke better than anyone I’ve known. For instance, one time years ago I was complaining to Grandaddy about someone’s behavior, hoping he’d chime in and validate that this person was in the wrong. But, I remember it as clearly as if it happened yesterday, Grandaddy smiling and kindly saying “you know, it’s a good thing that neither you nor I have any personal problems.” His tone was not sarcastic, but poignant. He was right, I didn’t have a right to criticize the person I was criticizing. To this day, I often hear those words in the back of my mind when I think about criticizing folks — too often, I don’t heed them.

I think the last post touched a nerve — it seems that many of you guys have had difficulty breaking into the precision GPS market. If you’re not yet inspired to take the precision RTK GPS plunge, just wait — we’re going to have some fun in upcoming posts pitting the ComNav K501G in an Epic Showdown with some of the competitors we mentioned last time. I don’t know about you, but I can hardly wait.

Rover 2

We now present the Rover 2 build as a photo essay: