Direct Drive Four Motor System

[Matt Adams]

Originally Posted by Jason Kixmiller

Actions speak louder than words (said or typed) and a winning season speaks for itself. Whether you custom built every piece or assembled from "the shelf," having a successful season allows your performance to speak for you. Many teams have great features/machines, but arguing about the best is in many ways missing the point.

Back to the true topic at hand, I have a question:

Is it feasible for the 4 independent motors to be electronically controlled/programmed to have the same output speed?

Absolutely, but it requires feedback, such as an encoder, resolver, tachometer, and a good controls algorithm, especially to control the speed of four different axis. However, I'll ask another question for you:

It is worthwhile to electronically control the speed of each individual tire in a closed loop fashion on FIRST robots?

This is an interesting question, and there are a few distinct and divided groups on these forums about drive wheel feedback / controls in general. There are some very adamant people who swear up and down that closed loop control (position and/or velocity feedback on the wheel) is critical, others who deem the resources / benefit are too small, and still others who ask, "Closed loop? What's that?" I consider myself to be in the 2nd group.

My thoughts are merely, "Can't a good driver with some practice on the robot before it is shipped 'aim' an open loop drive system 'straight enough'?" The advantage of a closed loop system is pretty nice if you have two pairs of motors controlling two wheels on each side - you can essentially ensure that when you push both joysticks full forward, that the robot will move straight and not veer off by say up to 10 degrees or so as it might in an open loop system. Driving is more intuitive, this is a very good thing.

Correct me if I'm wrong, but this thread has sort of now technically spun into a suggestion about a 4 wheel robot with independently driven wheels, each with an independent gear box, as well as a closed loop feedback setup on each axis to ensure that the wheels all spin at the same rate.

Others may disagree, but my opinion about this can be said in a single word: Overkill.

Matt

[Veselin Kolev]

Back to the main topic,

Both the times I have done direct drive to four wheels with 4 motors, I used a very similar setup.

I had a gear bolted to the wheel and the wheels rolled freely on bearings and shafts. The gearing was as follows: small gear on the CIM output mates with a larger gear, which is on the same shaft as a smaller gear. The smaller gear then mates with the larger gear that is bolted to the wheel. EX: CIM speed of 5342 x two 12:40 reductions = 480 rpm = about 12 feet per second on 6" wheels.

The reason for the two reductions being the same was so the first gear (on the CIM, 12 tooth) and the last gear (on the wheel, 40 tooth) could be on the same rotational axis. This proved for a very small drive train, and if you mess around with the shafts and gears, you can pull it off with very few parts, meaning less fabrication.

My reason for using a gear on the wheel itself is so it can act as the final reduction, instead of having extra bearings and shafts to have that reduction inside the transmission, only to run a shaft out to the wheel.

I really like the 4 wheel 4 motor independent direct drive system, it doesnt clutter your robot with drive shafts or chains, it makes everything modular. Also, fabrication is a blast, you're making 4 of the same part for the drive trains, and the gear reductions arent that complex or anything. No CNC required (unless you have interesting cheeseholes) and I would say this is a really reliable system, very low maintainance if done right.

The only way you can go wrong is if you use a face width that is too small for the gears. Then you start snapping teeth. I used .375 on all the gears, it worked quite fine. Also, it is a must to have your wheel on stable bearings and shafts, because if the gear on the wheel and the gear on the transmission misalign, the teeth get worn down or they snap. And the last rule of thumb, use loctite! When direct driving like this, there tends to be more vibrations through the drive train (it is usually dulled by the chain in chained transmissions). These vibrations make set screws and bolts back out and your drive train to misalign.. etc. Keep your drive trains screwed together!

[Jack Jones]

Is it worthwhile to electronically control the speed of each individual tire in a closed loop fashion on FIRST robots?

In my opinion, answer is: Maybe! But before you go piling on a whole bunch of complexity, you should try to design something simple that’ll go straight all on it’s own – build it – test it – if it works, you’re done.

That’s what we did with the above example of our experience, as Collin was asking, with four-wheel direct drive. It was a low risk experiment for us. The CIMs were known to be unbiased, and they gave us four. The parts were water-cut from scrap in about two hours – including the welding jig, which made for 30 minutes to weld. The DeWalts cost $21 each. The wheels cost $17. So, with a $200 investment, the kit frame, and some assembly, we had it on the ground in two days. It went straight.

OTOH, suppose that this year’s game has an autonomous that is worth doing and requires precise positioning… Who knows! Except that one thing we’re sure is we’ll try to keep it as simple as we can.

[Alan Anderson]

Originally Posted by Matt Adams

My thoughts are merely, "Can't a good driver with some practice on the robot before it is shipped 'aim' an open loop drive system 'straight enough'?" The advantage of a closed loop system is pretty nice if you have two pairs of motors controlling two wheels on each side - you can essentially ensure that when you push both joysticks full forward, that the robot will move straight and not veer off by say up to 10 degrees or so as it might in an open loop system. Driving is more intuitive, this is a very good thing.

You're missing something relatively important in your thinking here. Yes, a good driver can effectively be the element that closes the loop, but only when there is a driver at the controls. However, a reliable autonomous routine pretty much requires closed loop software control of motor speed.

[Billfred]

Originally Posted by Alan Anderson

You're missing something relatively important in your thinking here. Yes, a good driver can effectively be the element that closes the loop, but only when there is a driver at the controls. However, a reliable autonomous routine pretty much requires closed loop software control of motor speed.

I'd qualify that a bit with "an autonomous routine that has a decent chance of screwing up."

If you're just knocking off the hanging tetra, you can go open-loop (as the only thing you have to do is drive forward a smidge and raise an arm). There's very little chance of this screwing up, as you're not going far, and the hanging tetra is a big target.

Now consider the 10-point ball from FIRST Frenzy. Tiny target, really far away. Break out the encoders, folks.

[Jack Jones]

Originally Posted by Alan Anderson

... However, a reliable autonomous routine pretty much requires closed loop software control of motor speed.

Please don't tell our robot. We took the encoders off the drive wheels and mounted them on two boggie wheels about 6 inches each side of dead center - got better distance measurement that way. Direction was not a problem. Ask the folks at Rah Cha Cha or Novi - it would cap the center side goal dead solid perfect, or, with the gyro telling it when to stop turning, it could snatch a tetra from the autoloader and hold it over the center goal. The motor control was binary - on, then off.