After you solve the electronic, timing and mechanical issues, the big killer becomes one of traction.

For a tornado-drive to work by varying the speed of the drivetrain, the wheels have to be able to alter their rpm significantly within 1 revolution of the robot chassis.

How many wheel revolutions this equals is determined by the diameter of the wheels, but its not many, so you are asking your tyres to maintain enough grip to be able to exert a significant amount of force, whilst being scrubbed sideways across the floor due to the translational movement.

For the revheads out there, its roughly analagous to being totally sideways in a car, with the tyres sliding, then stomping on the accelerator and expecting to be able to accelerate forwards suddenly.

To not scrub the tyres badly, you have to be able to stop a wheel, *reverse*, stop and go forwards again within about 30 degrees of the chassis rotation. Of course the faster you want your chassis to spin, the less time you have to do this double-reverse and you quickly reach the point where again the tires just cant brake and accelerate that fast without slipping.

I believe this is why noone has made a tornado drive that has seemed to work that well so far.

Stepper motors may help the electromechanical problems, but I havent heard any ideas on how to solve the traction problem.

The most promising approaches seem to be those that lift the tyres off the ground or alter their toe-in/toe-out cyclically or a technique like that,

Hmm, I just had a thought. has anyone tried to do a tornado-drive / melty brain system using omni-wheels ?
_________________
Great minds discuss ideas. Average minds discuss events. Small minds discuss people

you don't need to reverse the wheels direction.
movement comes from a variation in speed.
small variation = small top speed.

cyclonebot would not have worked at all if it had to reverse wheel direction n hundred times a minute. I think cyclonebots major killer was over thinking the problem. They forgot they had a human in the loop to take up the slack.
A simple wheel rotation counter would "get you close" to being able to determine a direction and provided the errors were smooth (ie not bouncing around the place) a human could fix it and make it drive.

Stepper motors would make for sucky drive i'd think, they are generally quite low speed, decent torque though.
_________________
Mechanical engineers build weapons, civil engineers build targets

I dont think I've actually seen any evidence of cyclone bot translating in any controlled fashion. Its supposed to be able to, but has anyone seen it do so ?

To translate properly, it seems to me that you *do* need to reverse the wheels direction.

Otherwise, all you are doing is changing the diameter of the circle that each wheel is travelling in, but the retreating wheel is still going the wrong way, so the only way to have the bot going forwards while the retreating wheel is going backwards is for it to slip.
_________________
Great minds discuss ideas. Average minds discuss events. Small minds discuss people

I'd think about it from the point of view that the wheels are going in the direction of the bot's travel, but the bot's rotation is caused by the difference in wheel speeds.
So, if the rotation of the bot is faster than the translational speed, the wheels will all be going in the same direction, but slower when they're the retreating wheel.

If you wanted to rotate at a low speed while translating fast you'd have to reverse the wheel directions each rotation.
_________________
The Engine Whisperer - fixer of things

I have a video of the test models of Cyclone bot. It did translate properly, but it only need 300ish RPM.

Y-Pout is the only translational spinner that worked effectively, but it suffered from massive battery haemorraging on impacts. The Plague sorta worked, but it cheated...

To make a robot translate, you don't need to reverse any of the wheels. It's about speeding one wheel up faster then the other at precise points, Offset can be spun and translated across the arena with some quick finger movements from the controls, it's just hard to make it translate perfectly (Human's tend to err). I think Jakes idea is probably most simplest way of doing it, using encoders and counting the wheel revolutions.

I spent a long time looking at translational spinners and to make them hardcore and more effective then other spinners, you have to spin them over the 1000-1200rpm mark, which is bloody hard to do and maintain translation...
_________________
https://www.halfdonethings.com/