My team has begun using some of the WPILIb path planning tools and have had okay success so far.
We have a 4 falcon 500 wcd different drive system. Pretty standard.
Sometimes I’ve noticed that when reversing a trajectory (such as driving to get a ball, and trying to reverse back to avoid losing time).
I’ve noticed different results. I would hope that since I’m just going to do the path I did in reverse it would be fairly recreatable. Often it will be slightly different. 6 inches short… a little off to the left or right
Is this just all a function of characterization? We can use a PID loop to get an arm or turret to a setpoint with very precise accuracy, but wasn’t sure if that was realistic for a full ~120 pound robot.
How accurate is your auto driving?
We use all the standard controls…ramsete controller…falcon 500s and encoders… the kit of parts gyro
From when I tested our Ramsete I remember it being like a few centimeters off (I think ~5cm). I would chalk it up to tuning and gear ratio, as the bigger the gear the ratio the more accuracy the encoders will have.
I’ve never used Ramsete, so I can’t comment on how accurate it was for me. However, it’s really important to understand whether the robot is exactly where it thinks it should be, but it isn’t where it should be. Or whether it knows it’s off from where it should be but isn’t tuned appropriately to get there.
This could be from a variety of things, from wheel scrub to inaccurate gyro angle or motor counts that are off. From my experience if you follow the characterization procedures(and enter all feedforward constants + PD values) you can get really accurate paths(±1-2cm per meter driven). I would attempt running with a disabled ramsete and see if you have any better behavior. I often start by verifying all of my individual components. I make sure that the encoders and gyros are returning nominal values, then I run a straight path, a short curved path, measure, diagnose any bugs, and then I’m on my way. That being said, when running very tight quarters paths, tank drives are prone to scrubbing and you lose a lot of valuable data when that happens, I would recommend making very large arcs when testing autos until you understand more physically, the characteristics of your drivetrain. I’ve also found that after lots of hard teleop driving, our drop center is accentuated by uneven wheel wear. Once wheels are worn in more, scrubbing becomes less prominent and more complex paths are achievable. Again, all in my own experience, results may vary… wildly.
