Autonomous ATV

For Team Overbot, we used a Maxon servomotor with a planetary gearhead, coupled directly to the steering shaft via a shock-absorbing coupling. The steering wheel was removed. The motor had an HP encoder.

On the steering shaft, we mounted a circular through-shaft pot/encoder of the type used on GM cars. This produced encoder signals at 1 degree intervals, plus a signal from a pot for absolute positioning.

Control was provided by a Galil DMC-1416 motor controller, which also read the encoders and pots, and the auto/man switch. At power-up, the controller read the pots and centered the steering.

The steering wheel became a purely electrical input device, with an identical pot/encoder. This allowed manual driving with electrical power steering.

Power came from a 3KW Generac generator driving Vicor

Today, the Roboteq controller is a good choice. That wasn't available when we started. Get the one with the encoder input, and connect that to the motor's encoder. You'll need a second sensor of some kind to home the steering. A limit or home switch is sufficient. A useful trick is to have a home switch or sensor that will detect if the wheels are pointed straight ahead, plus a second switch to tell you whether the wheels are pointed right or left. Then you can home to center under program control. We used the through-shaft pot for that.

Incidentally, we steered the Overbot with less than

John Nagle

"John Nagle" wrote

Hi John, I'm very happy you could respond to my post, since my project is very similar to yours and you have certainly gone through the problems I'm having and the ones that I will have.

I was looking at the Maxon line of products and although I particularly like their products (very high quality), I couldn't find one that would provide enough power. Well, I will elaborate more further, as there are some interesting parts of your answer to be commented on. I had not though about shock-absorbing coupling, could you explain what sort did you use? I had selected US Digital encoders, because they are cheap (around $70). I didn't know that HP manufactured encoders, where did you buy them from?

I'm gonna look for it. I was planning on using the encoder on the gearhead output, and since the pulleys are 1:1, that should do fine don't you think?

Yes, we're going to remore the handlebar as well. The roboteq controller has inputs for RS232, analog joystick and RC RF receiver, so we can implement manual local control with a joystick.

That's interesting, and I saw a couple of entrants on the last darpa gc using generators. Why is that? Why not to use extra batteries and an alternator to recharge the batteries?

Yes, I was reading the roboteq user manual and they have a similar suggestion

I measured 45Nm torque at the steering column to turn from lock to lock on asphalt, I wonder if that's because the polaris ranger has a rack and pinion steering system, while in my case the steering column drives the steering rods directly to the wheels. Well, I guess total power should still be less than in your case, given that supposedly there is more friction to overcome.

Thanks for sharing your knowledge with us

Cheers

Padu

Pretty cool. Its refreshing to see someone who is spending some real money on this sort of thing. Got any more pics?

I would suggest that you use an electric motor attached to the drive train to allow you to implement a more simple acceleration and braking system at first. If your max speed in this mode is a few MPH then you can more safely debug your autonomous software. Once your software is working reliably you can switch into gas powered crazy mode without worrying about small software bugs creating a hole in the side of your house.

I attached a PC to a riding mower frame that was converted to electric power (RC airplane starter motor for drive train, laser printer motor for steering, two 100 amp RC motor drivers and a MiniSSC/RS232 servo controller) many years ago and found it is very handy to be able to chase a robot down on foot when something goes wrong.

I also found it difficult to debug something that can't operate indoors, so I am in the process of switching to a "Laptop on wheels" approach.

-howy

The encoder, the little black thing atop the motor, is an option from Maxon.

Below the motor is a spider coupling from Stock Drive Products. This provides some rotational shock absorbtion and allows slight misalignment, so that when a wheel hits a curb and applies forces big enough to change the steering direction, the gears in the motor drive aren't stripped by the shock load. Instead, some of the shock squeezes the rubber spider in the coupling for the first few milliseconds, allowing time for the gearbox to spin up the motor and force it to move against motor power.

You need to calculate the stresses and inertias for this to work right. Dr. Celia Oakley did that for us. When we did hit a rock or a curb, nothing broke.

The motors you're looking at seem OK, as does the controller. You have a belt drive, so you don't need the spider coupling. We had to couple two steel shafts end to end, so unless we put something in there, all shock loads would hit the motor.

The main encoder should be near the motor, not isolated from it by anything with flexibility, or you'll have trouble stabilizing the control loop.

Our experience was that manual driving with a joystick really sucked. We tried. With a joystick, everybody seriously oversteers. Just like R/C cars. We had to put in a steering wheel. Actually, we put the original steering wheel back on, so everything looked normal, but it was purely an electrical input device.

Our steering wheel had no force feedback; you could turn it fast and wait for the motor drive to catch up. This took some getting used to, but was tolerable.

Because the Polaris Ranger has a really wimpy alternator, and it's built into the engine, so you can't upgrade it.

Our lock-to-lock steering time was four seconds. That's reasonable for the Jeep-sized Ranger, but a lively little ATV might use a faster drive. Or not; you might flip the thing.

Good luck. Feel free to contact me directly if you need more info.

Incidentally, the Overbot is now at U.C. Santa Cruz. We donated everything to them.

John Nagle

Thanks for the links. It was very fruitful for me. Very good work you guys did.

I was afraid you would say something like this. I'm even more afraid to look at how electricity is generated by my ATV. With a Polaris like yours, it is simple to mount a generator since there is enough room for that. On a small

Interesting, that's almost one order of magnitude slower than what I'm planning. I wonder if such fast steering is really necessary.

Thanks John, believe me I will.

DAMMIT! Why didn't we get to be as lucky? :-) By the way, where did you guys would test your bot?

Cheers

Padu

Re power source:

We had more electrical power than we needed. We installed a Generac

Probably not. We went for speed on the brake actuator, but not the steering. If you're going to do super-fast steering, you're going to have to have all the smarts to understand the vehicle dynamics, or you will roll the vehicle.

Testing was tough, here in Silicon Valley, which has little suitable open space.

Team Overbot was in an industrial start-up space (1600 square feet) in Redwood City, CA, in the Redwood Junction industrial park. We went to considerable effort to rent a paved, fenced half-acre parking lot, where we were sure that if something went seriously wrong, no one would be injured.

Once things worked reliably in that small space, we arranged to get the use of a Sun Microsystems parking lot intended for some office buildings that were never built. That was about ten acres, but it was full of lights and traffic islands.

Then we did off-road testing at some horse ranches nearby. (I own a horse, so I have good connections in the horse community).

We had an industrial-grade emergency stop radio, an E-stop relay box which slammed on the brakes if a 125ms timer wasn't reset by the computers or the E-stop radio link was lost, and a general commercial liability insurance policy for several million dollars.

John Nagle