Hmm the end of the weekend has come too soon!
Today I tied the ultrasound sensors into a basic maze navigation system: in short, go forwards until within x of a wall, then turn on the spot towards the greater distance of left or right.
A difficulty I've been having is that my motors and wheels are just to beefy! The wheels are soft, rubbery 42mm wide, 120mm diameter affairs and get a Lot of grip. The motors can deliver 4 Nm each. And in the middle there's a timing belt...which is not up to the job of transmitting this much torque! An especial difficulty is encountered during neutral turns. Here, one side drives forward, the other back, and the rover spins on the spot. However due to the massive wheel-surface friction there's a lot of resistance to slipping the wheels laterally. The motors max out, the wheels stick and a staccato burst of teeth slipping follows.
This has the main issue that when under manual control I can rectify this, typically by avoiding true neutral steer and I can always resort to a 3 point turn. However, for the autonomy system such nuanced control isn't possible (or at least is a lot more complex!) Ideally I want it to try to spin on the spot, and actually spin!
So my issues were the wheels are grippy laterally, and the timing belts too slippy!
Firstly the timing belts: its too late for a major redesign now so something has to be done! Effectively, they transmit torque via friction, but instead of relying entirely on surface friction the teeth provide a cogging effect. Although not if so slack they can slip! So we need to increase tension to prevent the teeth jumping. Also remember that standard friction is Fr = u x R. Changing the coeff of friction is possible, but not easy within timeframe and cost. However that R, the normal reaction force, is supplied by the belt tension. Okay so I need a way to increase tension. Unfortunately I designed out the tensioner arms the Mk1 rover had because I thought I wouldn't need them this lime (oops.) But I had a cunning plan: instead of adding tension to the belts, add normal force at the pulleys (ie the reason I want tension anyway.) How?
Rubber bands!
Each pulley & belt pair now has 3 wide rubber bands stretched over the top of it. Now belt slipping is far reduced!
Excellent!
But what about that other option - decreasing wheel friction? Well I had another cunning plan: talc! The wheels have very pronounced tread so if I put talc just on the high points it would reduce friction for sideways slip, but the tread walls and the base of the tread would still be the normal sticky rubber and so could still grip obstacles for climbing.
As an aside it turns out baby powder is now no long talc, but rather cornstarch! I don't know if this would work, but not wanting to take any chances led me to a rather entertaining trek around supermarkets and pharmacies where I kept getting into the particulars of chemical formulas with store assistants. It was all made worse by my being rather sweaty having just come from the gym, but I couldn't decide if it would be more alarming to the store staff that I wanted it to build a robot than the thought of it combatting sweaty areas !
Anyway, having found talc, it worked a treat on the wheels and now the robot is pirouetting like dream. Now to actually test the autonomy once its all reassembled!
