[Dprglist] straight lines

[David P. Anderson]

Hi Paul,

Well, yes, or perhaps a qualified yes.  Since there are several thoughts 
here let me answer each as best I can.

On 1/14/22 8:39 PM, Paul Bouchier via DPRGlist wrote
> Thanks for the videos David. As always, your robots run very precisely 
> and deliberately.
>
> One question about the right wheel speed overshoot after hitthing the 
> mat: I remember you showing some plots of your PID controller response 
> to a step input a bit before Christmas, and showing that there was no 
> overshoot when accelerating (IIRC), and what you thought might be 
> overshoot when decelerating was actually a kind of elastic reaction to 
> the negative voltage being used to slow the robot (IIRC), akin to what 
> happened when you tried to manually turn the wheel. I may not have 
> understood it fully at the time, but

Close.  Let me emphasize first that the test before Christmas to attempt 
to determine if the PID output actually goes negative was a very special 
case, abrupt deceleration from full speed to zero, that never actually 
happens in normal robot operation.   It certainly is not happening in 
these videos.

Second, when I tried to manually turn the wheel with the velocity set to 
zero, the force increases as I try to turn the wheel further.  The 
amount of correction is a function of the distance of the wheel from the 
set point,  not how long it was disturbed. So I can turn the wheel a 
small amount manually, but as it turns further it gets harder and harder 
until my hand can no longer overcome the force of the motor.

> 1) isn't such an elastic reaction actually overshoot inasmuch as 
> velocity doesn't smoothly ramp to zero?
But the velocity does smoothly ramp down to zero in normal operation, as 
I illustrated at the time.   Again, the PID output going negative, that 
which you refer to as an elastic reaction, only occurs in the extreme 
testing scenario, and does not occur in normal operation.
> 2) in the video from yesterday, do you know whether the acceleration & 
> "catch-up" experienced by the wheel that hit the mat was solely due to 
> the PID loop building up "pressure" (i.e. an overshoot), or is it in 
> part or whole a result of a change in setpoint caused by the IMU 
> detecting rotation and correcting for it?
The navigation behavior on the robot was disabled for these videos.   If 
enabled, what one would see is not only the robot attempting to maintain 
a straight line, but also attempting to always steer toward the 
waypoint, whether that requires a straight line of some other maneuver.  
That is clearly (well at least to me) not happening in these videos.
>
> Your robots are precise enough that you can see subtle effects but 
> their cause isn't always clear - just trying to understand it.

Yeah, not always clear to me either!

cheers!

dpa