pic: Cyber Blue Swerve Chassis - Rear View

[AustinSchuh]

Originally Posted by reversed_rocker

when we built our crab drive, one problem we had was that the motor would over shoot the intended position, then try to swing back, and the process repeats.

This is a classic controls problem. You guys have implemented a "P" controller. Adding a derivative term will help, and make it a "PD" controller. One way to tune the system is to increase Kp until it gets pretty close to oscillating constantly, and then increase Kd until it doesn't overshoot any more. If at any point it starts acting "funny", decrease the constant some. One thing that really helps is to get back the position data at each point in time and then plot it.

I believe that there are some white papers around, but until I finally got the chance to take a controls class and talk with someone who does controls for a living, I had a little trouble with tuning the loops too.

I've used the following successfully many times in the past.

power = Kp * error + Kd * (error - last_error)

[reversed_rocker]

well its not so much an overshooting that we had problems with, it was that the motors would start vibrating because they were trying to switch directions so fast. it seemed harmless at first until the motors started getting hot. if you multiply the delta angle by a constant (where constant < 1) like i was describing it should give you a little bit better response

[Steve Warner]

Actually, to correct Collin a little. we are currently using the P term of the PID control with very little or no overshoot. However, we do have a maximum and minimum speed and a small deadband around the target angle. This seems to work pretty well on the bench but we haven't tried it on the floor yet. I know it will be different on the floor but I'm wondering if we will see a difference when the robot is standing still and when it is moving.

[Matt Goelz]

we got an official weight tonight. The robot with no battery and no bumpers, weighs 77 pounds.

I am not sure if this was already posted here, but we also have a video up on youtube of our first practice session with the robot.

that can be seen here

Matt G.