Tanto 2

[jamesb]

Tanto 2

ps... this is a hint to ask me for advice on making efficient traction control.

[fluppet]

Tanto 2

Ill bite, James (especially since some of my more obscure designs would be a bit lacking in the physical traction department). Care to start a discussion/advice session in hints & tips? (Hmm, we dont have a section thats for nonspecific weights; unless admin are so inclined, I guess we default to heavyweight.)

--
Fluppet

[fwsorg]

Tanto 2

Ed, Ive had this conversation with you before ...
... I do not design Robot wars electronics which rely on sensitive Chips- I do have a modicum of common sense. As I have explained to you before, should any of the modules (for example the accelerometer) stop working, the 2 processors will simply work without them. Should 1 of the processors fail, the other one will take over its job. The whole speedos are designed for things to go wrong! Short of an axe chopping the thing in half, weve tried to think of every eventuality!

[fwsorg]

Tanto 2

Andrew- Basics of traction control (and this will include slip angles :-

Contrary to popular beleif, a robot is not generating the most force if its wheels are spinning- the most traction from a wheel is obtained when the wheel is only just on the verge of slipping. So if you want to most traction when pushing another machine, you want to reduce the power of the motor (assuming the wheels have been slipping) untill the wheels are only just about to spin.

A slip angle is the angle at which the surface of the wheel will just break contact with the surface that it is on. The slip angle therefore depends on the surface youre running on, the amount of pressure on the wheel etc. For most traction, you must apply power to your wheels so that they only just begin to slip- this is the slip angle. I invite james to explain this far better than I because Im not very good at explaining such concepts, and hes had a lot more experiance with this kind of thing that I have.

[jamesb]

Tanto 2

yeah, here is the secret, Ive been sitting on it for a few years, but as Im not going to do traction control, I might as well let the cat out of the bag now.

To achieve efficient traction control you need the following.

a realtime feedback loop, measuring wheelspeed (or motorspeed), lateral g, groundspeed.

you also need data about your wheels on the required surface, this can be simulated on a rig, but finetunin must be built in.

Right, the way this works.....

in a straight line, the robot will accellerate and decelerate when the tyres are operating within their maximum effective slip angle.

to find out the value of this (this also explains what a slip angle is) you need to run the wheel on a rig. this rig is set to measure the torque of the motor, and the torque on a slave shaft which simulates the floor of the arena. This way you can measure grip of the tyres.

accellerating the motor should accellerate the slave shaft at the same rate. If the slave shaft accelerates slower than the motor, you have wheelspin. A torque comparrison gives you a graph of how much wheelspin you get.

you will find that, from when the speeds first vary (ie, wheelspin first occurs) to a noticable difference in torque, there is a gentle squeeling. This is a tell tale that the tyre is working at its hardest, and is within its slip angle (the theory being, the wheel turns, say 367 degrees for the slave shafts 360, so it is an angle of 7 degrees different caused by the slipping.

so what does this mean?

if you look at the graph, you will see that the accelleration did not drop for this area of slip. Beyond about 7 degrees (this is an example for a go kart tyre) the torque on the slave drops suddenly. this is where you need the traction control. Regualr traction control cuts in as soon as the wheel speed is different. this ignores this area of slip, and so automatically looses it.

by having groundspeed, and motor speed, you can program the traction control to allow for the slip angle, as derived from your rig tests, and so maximize your available traction. This works in deceleration to, as for example F1 abs was working within slipangles, unlike car ABS.

to maximize lateral grip, you need sensitive g-sensors, but best avoid thse, as they will get confused when storm 2 hits you.

there we go, a nice description of why previos traction control didnt provide enough of an advantage to make it viable, it cut off the bit where most grip was produced.

oh, before everyone starts looking at this, think about one more thing.
Tornado pushing on the side of... um... Smidsy, for example.
Cant move him eh, wheels are not gripping? traction control would not help here, what would help is spinning the wheels up for a second, which covers the floor in tyre rubber, then push again and voila, more grip.

traction control, Ill never use it.

[craig_anto]

Tanto 2

I cant honestly say Im interested in putting on a weapon but Im sure me and Eddy will argue and Ill end up with some kind of weapon. Its gonna have spikes it that counts?
Traction control well Im not saying anything as too what Ive got in mind.
The angle is too steep? its 45 degree I think thats perfectly ok. When Tanto catches any edge on the wedge it tends to turn it over.

[matty_haka]

Tanto 2

Craig, no I wasnt at Brighton! Seeing as Im only 14, getting anyone to drive me anywher is abot as easy as threading a piece of rope through the eye of a needle, so Iam impressed that you managed to throw something out the ring, though without seeming rude (It could in fact be a rare compliment! :wink I wonder whether it was the wedge that enabled that or the sheer brute force of the thing, aided by the shape!
As for a weapon, Why not have an Aggrobot3 Style wheel underneatth, that way you can raise the nose, and if it stops working, youve still got the wedge and brute force to carry you through.

Oh, and on a final note, is James the first person to become an advanced roboteer?! Congratulations if he is, and alos is ther a higher rank than that?!

[craig_anto]

Tanto 2

just updated the tanto 2 bit
its not a good site i know but we do the best with what we have

[fluppet]

Tanto 2

Thanks James. I was aware that maximum acceleration usually happens with some slippage, but hadnt seen it expressed in those terms before. I presume a fair bit can be done just by sticking a strain gauge on the driveshaft and seeing how much force you *should* be applying to it compared to the change in drive speed? Im a bit wary of the idea of predicting how much traction ought to be available allowing for heating/wearing tyres, fairy liquid and so on.

Other thoughts - can anything be done with the mechanism from an optical mouse to measure floor speed? (If some optics cant scale up, itd only help for antweights).

However, this is off-topic; perhaps a move to tip swa^H^H^H^H^H^H^Hhints and tips might be better if people want to discuss it further.

Several of my designs (sorry, vapour builder here) would benefit from having some intelligence on board; once the programmability is there I may as well add some engine management. :-)

(Speaking of which, whoever came up with the traction control for the Focus didnt presume that you want to accelerate if your wheels are spinning; kills the engine revs until youre out of the torque band, not helpful, better to balance the clutch and keep the revs - if not torque - high.)

--
Fluppet

[fwsorg]

Tanto 2

Andrew- Ive been playing with optical mouse technology for that very reason. It was useless at 20mph, but certainly for ants it would do fine. A small wheel with an odometer seems to be the best option.

[fluppet]

Tanto 2

Cool - I look forward to seeing how well it works. I suspect the technique should scale up to higher weight/speeds if a bonus lens was stuck in front of the sensor (think fisheye), but that would be a bit fiddly to set up and might be too damage prone.

If you get it working on an antweight, Id be interested in giving it a go, though, if nobody else tries it first (or they do, and it works). Might be helpful in my plans for a self-built robotic vacuum cleaner, too (the ones you can buy cant do stairs, and I like a challenge).

--
Fluppet

[jim]

Tanto 2

Ive been looking at optical mice too, in fact I have obtained some code to interface a PS/2 optical mouse to a PIC, although I havent actually tried it yet. My research indicates that the highest speed they would work at is about 2mph. I would be interested to know how you get on with them.

[fwsorg]

Tanto 2

I think traction control on an ant is a little bit silly :-) Would be fun tho. The prob with ants is that the weight limit isnt quite enough to stick in our own pic based circuits and PCBs, and start playing with things like that, unless youre willing to go down the SMT route which I most certainly am not :-)

[fwsorg]

Tanto 2

Having said the above, I am now determined to build traction control for an ant The odometers from a ball mouse for the wheels, the optical sensor from another mouse for the real time feedback, and a PIC circit as small as i can get it, which will sit in-line between the Rx and servos (itll do the mixing as well). Im doing it to see if i can, not to win any competitions:-)

[fluppet]

Tanto 2

I approve - and Ill watch out for it! Ive always thought robot combat should be about interesting designs, rather than the tried and tested. (Hence Im a fan of Tip Top, in spite of the drive mechanism blatantly being a bit of a daft idea.)

--
Fluppet