Closed edufford closed 6 years ago
I was also surprised by the quick movement of the steering wheel at the start of the video, although it is an area of tight turns. I looked at the spreadsheet and also checked the difference between values, which as expected was generally less - suggesting smoother variations in steering.
Unfortunately, I don’t want to make any assumptions about the curvature of the waypoints that Udacity sets for the test or where the car might start from so I’m not sure how sharply the car needs to be able to steer to get on track. Since they didn’t make any note about the steering in the first test, I guess it wasn’t too fast with the base tuning but I think this small change will help a bit without changing the response too much since we know it was able to keep the line with the previous value.
In the video of the v1.0 site test, the steering wheel has some rapid rotation at the beginning and end of the test. Increasing the steering LPF tau can help smooth out the rate of these large rotations, while still being able to track the overall steering movements needed during the drive.
See the attached chart showing the result of the dbw_test script run on the
1150531_2018-04-12-17-02-51.bag
data, showing the original steering command with tau 0.1 vs the proposed steering with tau 0.2:steering_tau_retune.xlsx