Closed griswaldbrooks closed 6 years ago
Results of first pass. https://youtu.be/CNS4LPSbZGY
It was trained using multiple PID settings, but the majority were detuned settings so that the robot would encounter recovery situations but it seems like there were not enough "good" situations and the robot "thinks" that oscillating is what it is supposed to be doing.
Trained with multiple pid settings, but the majority were detuned constants, so now the robot seems to think it should drive in a wonky way.
Run video: https://youtu.be/CNS4LPSbZGY
Training data https://s3-us-west-1.amazonaws.com/kittcar-data/sim/pid1.tar.gz https://s3-us-west-1.amazonaws.com/kittcar-data/sim/pid2.tar.gz https://s3-us-west-1.amazonaws.com/kittcar-data/sim/pid3.tar.gz
Using pid mode in simulator, train a network.
Need to use multiple pid gains to induced situations where recovery is required.