You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In Leaderboard 2 the carla vehicle seemingly reacts differently to the input steering values. For once it looks like, they inverted the steer [-1 ; 1] in the carla_control_cmd message to now steer right on negative values?? ( negative was "steer left" in Leaderboard1)
There is a PID in vehicle controller that controls the steering to smoothly instead of abruptly steer into the newly calculated steering_angle.
-> This PID is not tuned, and in fact that badly tuned, for leaderboard 2, that we can not drive with it.
-> Tuning this PID and inverting the steering for the carla comand message will hopefully finally get us to finally drive again in leaderboard 2.
Definition of Done
Tune Steering PID in vehicle controller, so that calculated target steering from steering Controllers (Stanley and PP) is well followed with at least 95% accuracy and no overshooting.
Fix the steering sign in __map_steering which also somehow tunes the PID inputvalue? why? 100% correctly steer left/right.
The text was updated successfully, but these errors were encountered:
We may do not need this PID since it only delays the steer input (which should be optimal from our steering controllers) so if at a given time/position a steer of 30° to the right is optimal, we should just drive at 30° right steering and not slowly go from 0 to 30°. This needs further testing as absolut steering inputs ARE POSSIBLE (tho not realistic but the simulation does not punish us for using jumpy steering inputs) BUT MAYBE the jumpy nature of this leads to more unstable trajectory following?
Disabled the PID in vehicle controller and also the strange tuning of the calculated Steering-Inputs from the Steering Controllers. This also removed the clipping of the carla-steering-singal to -0.75 ; 0.75 to now use the full range -1 ; 1.
Vehicle_Controller should only "map" the steering-Inputs (in rad) to [-1 ; 1] and invert them (since for the carla_steering-signal right and left are just inverted from what is calculated in both controllers).
This simplifies the understanding of the signalflow for steering alot, and allows proper tuning of PurePursuit and Stanley in the future.
Description
In Leaderboard 2 the carla vehicle seemingly reacts differently to the input steering values. For once it looks like, they inverted the steer [-1 ; 1] in the carla_control_cmd message to now steer right on negative values?? ( negative was "steer left" in Leaderboard1)
There is a PID in vehicle controller that controls the steering to smoothly instead of abruptly steer into the newly calculated steering_angle.
-> This PID is not tuned, and in fact that badly tuned, for leaderboard 2, that we can not drive with it.
-> Tuning this PID and inverting the steering for the carla comand message will hopefully finally get us to finally drive again in leaderboard 2.
Definition of Done
The text was updated successfully, but these errors were encountered: