Farm Weed Spraying Robot

Hi All,

I’m looking to try to build a robot that will navigate fields accurately while selectively spraying weeds.

The spraying part will be done via switches/relays/solenoids, and doesn’t need to be connected to the controller.

The chassis will be a converted 36v mobility scooter, which has some excellent components by the way.

I can fabricate and wire to a very high standard, and have access to good equipment including a reasonably serious 3D printer, but my programming/code skills are non existent.

I’m looking for the best plug and play option, and am considering the Pixhawk 2.1 with the RTK gps.

I’m seeing a second GPS module is required for the base station…is that a second here+, or is it a part that’s included in the set?

Does the base station/laptop have to be setup for the entire operation, or can it be switched off and packed up after kick off?

Is the Pixhawk gear worth the spend, or is there a better option that is simpler to use?

Basically what I want to do could be considered similar to an automated electric lawn mower project.

Would appreciate any advice.


The base has to be set up and running any time you want an RTK solution at the rover.

Pixhawk is a great controller, highly flexible and extensible. However, the learning curve is pretty steep. Unless you are trying to do something very unique, the pixhawk can probably handle it with no additional coding.


Thanks David,

I’ve really struggled with coding in the past…

I don’t imagine my needs are too unique. I’d like to set up “fences” on the pasture in mission planner, then have the robot travel the entire area in a grid pattern so as to cover every inch.

I’d like a return to home function, but through gates and lanes, not just as the crow flies.

At the risk of sounding silly, what do you mean by an “RTK solution at the rover”? By that do you mean the cm level of accuracy supposedly provided by the here+ gps module?

If so, can the base be many miles away from the rover, or does it have to be in the vicinity?

This is easily done with mission planner or most other ground control stations. Similarly, covering a grid is easily accomplished wtih mission planner. But if you are trying to go up and down rows, grid may not be your best option. You may have to create a “flight plan” selecting waypoints along the desired route. The rover will go straight line from waypoint to waypoint.

You will probably have to create a route with waypoints to do this.

No worries. Yes, the only way the rover will get cm level accuracy is if its receiving a correction. You can get a correction from your base station or from a CORs network if you have one in your area. Using you own base requires line of site between your base and rover. If you have some high terrain, put your base there. Most telemetry radios in the US are 915mhz, which really doesn’t bend around to terrain.

The other option is to use a CORs network, although I am not sure if that is an option with the HERE.

Once you get this figured out, can you make me a pixhawk powered lawn mover? Been thinking about that for awhile!


Thanks again,

So is there an accurate GPS module that works with the Pixhawk 2.1 to say 10cm accuracy but doesn’t need a base station in the area? What are my best options without a base station?

How about this:

Have you confirmed you have a CORs network in your area?

I can’t tell you what is best, I can only tell you what’s working for us. We are using a AstRX from septentrio. This is a great system that is plug and play with the pixhawk with cm level accuracy. However it’s quite expensive. For some information about other options you may want to read spilling the beans on RTK in this forum.


Thank you,

I think it’s best that I just start with the standard gps and see how I go even using this Pixhawk/Mission Planner first.

Obviously I don’t know enough to throw down big money on an RTK unit straight away, it would be better to have the thing built and rolling first.

The way things are going it seem that the more I wait for that part, the cheaper and better they get.

It does seem that we have no shortage of CORs stations around my area (Kempsey near Port Macquarie, up the coast from Sydney), so that does seem encouraging.

I did read that article by the way, and if we’re in agreement that the Pixhawk 2.1 is the right package for me I’ll start that way and upgrade later.

I really appreciate your advice thank you

I am a big fan of the pixhawk and ardupilot. I am still using the legacy pixhawk 1. Hard to accomdate the new form factor in my old builds, but space shouldn’t be a problem for you. I’ve never used rover, but like copter and plane, it has a great community.

Good luck,


Hi Neil
I am working on a couple of projects related to weeds, one ground based using Pixhawk and Mission Planner, the other is aerial detection of weeds. I am Sunshine Coast based north of Brisbane. Happy to share thoughts on all this.
Did you catch last Monday’s Catalyst program They are doing some very innovative work at ACFR in Sydney.

1 Like

Mate I’d love to swap notes with you.

I’ve PM’d you with some contact details. Please feel welcome to touch base.

I’ve not even started yet, it’s just an idea at this stage, but I have sourced a donor platform…

It’s the one and only ever completely Australian made mobility scooter. The company shut down well over 10 years ago.

It’s only 24v and has no suspension, but has twin direct drive motors on the rear wheels which apparently makes them much less prone to bogging and highly suitable for my running backwards with dynamic/skid steering idea.

I’m told there’s not a single Chinese part on the thing, it’s BHP steel, Preslite motors/gearbox and even (pre Chinese) Arlec onboard battery charging. Probably why they went down, too expensive to produce. Apparently they are highly prized for converting to a golf buggy.

What I like is that being Australian made, it’ll be all off the shelf components in standard metric sizes and minus some of the strange engineering practices you see out of China in their cheaper stuff. If I want to fit a new motor/wheel/whatever, there’s a better chance of sourcing something that’ll bolt straight up.

Haven’t got it in hand yet, but have settled on a price. The guy seems OK, so it should come off.

1 Like

Here’s some pics of my new ride:

All I can tell looking at these pics is that these motors are worm drive with brakes on the back.

I’ve read that’s a good thing as you can remove the brake assy and fit encoders for autonomous operation. Can anyone explain whether I need encoders, or whether I should keep these brakes?

I’m considering this version/clone of the Pixhawk as it’s getting good reviews for quality and comes in an ally case:

The Curtis motor controller fitted already drives both motors as one, so I’ll remove that and look at one of these for a skid steer operation:

Most likely this one:

I’ll just have to look at the rating on the existing controller first. If it’s 70A I’ll try the one above, if it’s 110A I’ll go with the next size up. I’d like to see if I can use 36v with these motors, in which case I’d be drawing less current and can use the cheaper controller?

As I said before I’ll run it in reverse, and let the (previously) front steering wheels move freely like a pair of joined castors for now.

Would love some input from you all with regards to what you see there.

I’ll pick it up in 2 days time.

Here’s the chassis/motors/steering with all the scooter stuff stripped off:

Now that I’ve removed the brake off the back of the motor, I’m straight away faced with an issue regarding steering. If I run it in reverse which I prefer to do for front wheel drive traction over pasture/field terrain, the steering arms don’t turn with corners like castors would. There’s a few ways I could go about it.

  1. Rear wheel drive (both motors as one) and front steering.

Really don’t want to do it this way, as the control and traction just won’t be satisfactory. Would be able to use the original Curtis controller and fit some sort of linear actuator or planetary geared motor on the steering rack.

  1. Front wheel skid steer and change to free spinning/individual castors on the back.

Some fabrication to frame, but no steering motor at the back. Bit concerned about how the castors would work over rougher country, and about the buggy “falling off its line” when driving across the side of slopes, especially loaded up.

  1. Front wheel skid steer and motorized rear steering combination.

Is this possible with Ardurover/Pixhawk? Certainly the most complicated for wiring and programming, but the control and traction would have to be the best of the options above over all terrain.

Could someone please tell me if it’s possible to configure it this way?

Got a little further along with it today.

Money’s a bit tight for hobbies right now, so I’m just aiming to get it to RC controlled stage for now.

Power goes through a 100A midi fuse, then an estop and onto a terminal block, which is where it stops for now until my motor controller arrives.

I’ve left room for the controller and the short (and ugly) red lead between the estop and the yellow terminal block is possibly temporary, allowing space for me to put an inline Mauch current sensor in its place if I go that way.

Cables are all 6 B&S with hydraulically crimped and heat shrinked lugs. About twice as thick as what came off it, but as I had all that gear laying around from my dual battery system install it seemed to be the way to go.

Batteries are rotated so as to keep the main power leads shorter, which was mentioned a few times in the motor controller manual and because it was much neater to link them in series this way. I also like the possibility of fitting larger capacity batteries if it all goes well.

Estop is right on the front of the machine which is possibly a risk of it getting bumped, but it just fitted so well that way that I went with it for now.

It’s got a way to go now, as the steering linear servo I want to use is expensive. I’ll get nice 5mm stainless allen cap bolts of the correct length to mount it all up neatly and order a Lemon/spektrum satellite receiver to control it with.

I’ll get it to the point of motor/throttle operation by RC first, then look at setting up the steering.

I said previously that I’d run it in reverse for the extra traction of front wheel drive, but I might set it up for normal rear wheel drive when moving it around with the radio and front wheel drive for slow autonomous operation over pasture.


Little further along…

My servo has arrived and I’m just playing with a mounting position while I wait for my 24/12v converter and a new receiver.

Some very rudimentary mods to the steering rack:

It should be noted that in normal centered position this servo will protrude some way out to the side, but nowhere near as wide as the weeding implement I’ll be adding.

While I’m happy with how this will fit up and work, I do regret not using a dual motor controller and skid steer as it actually would have cost less and have less moving parts, but this chassis already had a steering rack so it is what it is.

Once again more waiting for funds and parts, but soon I’ll have the slowest and ugliest RC car in the world.



Looking good! Nice to see the progress.

I was reading back further in this topic and I see you mentioned you wanted the return-to-launch to drive around gates etc. Recent versions of Rover include SmartRTL which probably does what you want.

Thanks very much for that, and yes that might save some issues.

As I’ve said previously, fabrication is my strong suit and programming code is not.

When I get to the autopilot part I’ll be here begging for help, but it’s the spendy part and funds are tight. I’m looking at the “Pixhack V3” variant, as I like the ally case and pin positions. My understanding is they all run the same software.

I have a Chinese RX and switch mode power reducer for the linear servo on the way (only available in 12v), and will have an update soon enough.

I do appreciate all help.

1 Like


Sounds good. Yes, any of the Pixhawk compatible boards should work.

In general we recommend that people buy from one of our partner’s store because they contribute to ArduPilot which helps us keep the lights on but it’s just a suggestion.

Just a little further along today…can only do what doesn’t cost money.

Temporarily hooked up some 12v power (off one side of the battery bank, not kosher) and a radio receiver, and I now have drive and steering as the linear servo is centered so I could mount it.

I’ve probably bought too slow a servo, thinking it would help with the steering “hunting” when operating autonomously. Went too heavy on the slow.

For the record I’m a little annoyed at myself for about this, as this servo cost 500AUD and I could have checked what I needed better before buying, especially as they have the speed I need at the same price.

At this stage I’ll run this just as a rover to learn Ardupilot/Mission Planner, as well as play with running an onboard honda inverter generator (I have the 1kva and the 2kva) as an onboard charger/range extender just to see how that works.


And a bit further along again…

I sprung for the faster servo, and at 4x the speed of the other one it’s still slow as hell!

I was concerned, but when I’ve limited the top speed on the radio back to what I think is about 5km/hr (my target speed will be 4-6kms/hr) the steering should be fine for the intended purpose, which is navigating perimeters and survey patterns in open paddocks.

While I certainly wouldn’t recommend this servo for a faster RV type rover, it’s certainly precise and pretty strong too. My 65kg teenage son stood on the thing while I drove it around with no problems whatsoever, in fact it steered better for the weight. I stood my 115kg carcase on there as well with controller in hand and same thing, didn’t even notice it.

No money to spend for a little while, but in the meantime I’ll cut off the round socket the seat slotted into and leave the rest of the post as a pedestal to mount a large white PVC junction box onto. This should give me a safe spot for the electronics and be a neat job as well.

I’d really like to see the wiring be a bit neater, and I’ll need to clamp down the batteries as well.

There’s no worries about it pulling a payload and traction, it pulls like a train.

I’ll be very interested to see run times, but I’m certainly a long way under the regular loads it would have been carrying as a scooter.


It’s looking good and definitely taking shape!