Quote:
Originally Posted by Ether
|
I would love to see this done well in competition. It's certainly possible in theory, but it has some challenges apart from those you get doing it with a true holonomic drive.
Team 1988 tried it in the "Overdrive" game with my not-quite-sufficient assistance. The robot was optimized for easy turning, but at the cost of straight line stability. In the end, it was indeed "a hoot" when it worked, but there were issues with control of angular momentum and with gyro limits that made it hard to drive in game situations. You could never quite guess which way the rotation transform thought it was going, and the gyro reset got used nearly as much as the stick.
We learned a lot and it seems like many of the issues could be managed with better mechanical design, so it would be fun to try again with that plus the improved gyros and closed loop motor controls we have now. We haven't gotten around to it because our drivetrain focus has been on learning to do holonomic right.
Field-relative control software for non holonomic drive trains must make decisions in the software that aren't issues with omni/mecanum drive trains. In particular, the tradeoff between changing heading and changing position. Holonomic drives have the same issue, but they're entirely taken care of by physics - the software doesn't have to specify the tradeoffs. I'm guessing a properly tuned program could approach equivalent performance if the software were carefully matched to the hardware, but it's not trivial to do, and it's on top of software that's more complex to begin with.
Either is correct, it's absolutely possible. But unless/until it's successfully done in real life the ability to implement field-relative controls must number among the advantages of holonomic drivetrains over others.