Hi,

as per usual I wanted to see what i have just learned in action. I have quickly recoded my A* algorythmn, which i showed off last week, to work "cell by cell" in order to get a non smooth path innately. In addition a safety distance to walls was added. The resulting path (green), is then smoothed (blue).

The car was configured to have a gaussian steering error with a variance of 20% of the current steering and a first order steering lag (PT1) with t=0.2 (95% of desired steering is reached after 5t = 1sec). Since the car has errors now, i can't drive the path exactly (like I did last week), instead the tools we have just learend about have to be used.

PID Settings were: kP: 0.1, kI:0.05, kD:3000. They are guessed (no Twiddle).

P only-VIDEO youtube.com as you can see the car goes mad really fast. At some point the overshoot is so big that it will continue to go in circles because the error stays greater than 0.

PID-VIDEO: youtube.com
It is interesting to see how the PID controller works. When the car is on the very top, before the last sharp turn, it drives paralell to the path. I have found out that this is caused by the integral value. While this error isn't good, it drives overall better with an I-Term. Either it's something you have to accept, or maybe my parameters are just horribly wrong :D

UPDATE: I uploaded the new version to the download page, futhermore I finally made a twiddle code which gave me better values for the pid control. The "run()" function of the twiddler simulated an entire run throught the course shown below (blue line), and summed up the suqare of all errors, just like the one we used in class.

PID-VIDEO2: youtube.com without PT1 lag, without any errors and with the values gained fromt he twiddle code. While the car stays on the path better on average, I have to say it would be a little bit shaky inside this car :P

Screenshot: