Car sometimes couldn’t stop in time for late detected lights

[edufford]

During Udacity simulator and site testing, the red light detection was late sometimes so the car judged that it was too late to slow down in time based on the estimated minimum stopping distance. Some improvement to the estimated stopping distance may be needed to allow more chance to stop for late traffic light detections.

[teeekay]

After running tests, it looks like we can't get much out of this. Shorter distances will result in higher deceleration rates, that can be magnified in dbw response. It looks like use of abs_max_jerk of 6 is suitable for calculating a stop distance that won't exceed jerk or acceleration limits of 10 in the simulator from speeds of 40 or 60 km/hr.

[edufford]

Ok, using 6 for max jerk limit sounds good to still keep some margin for actual decel rates during braking control. Will it help reduce the distance from 2.7 m/s at all?

[teeekay]

On last test it was about 1 m shorter. But that wouldn't have made the difference on the test.

[teeekay]

get_min_stopping_distance found shortest Distance of 3.387m to decelerate with start_jerk=-5.789, end_jerk=-3.579 ,from v=2.667, a=0.000 to v=1.250, a=-0.100 in 2.540s - Took 0.0002s to calc

[edufford]

Ok, every bit adds up though so all your improvements definitely help.

[teeekay]

with new improvements in PR #121 get_min_stopping_distance found shortest Distance of 2.130m to decelerate with start_jerk=-5.958, end_jerk=5.315 ,from v=2.667, a=0.000 to v=1.250, a=-0.100 in 1.088s - Took 0.0011s to calc

[edufford]

Improvements have been implemented, closing this issue.