17.6.3
Description
This PR fixes the accuracy of time calculation needed to zeroing acceleration in FollowTrajectoryAction.
With this fix, the action stops at the final waypoint more reliably - it more often meets all requirements: velocity = (0,0) and acceleration = (0,0) at the final waypoint.
No regression was confirmed here.
Abstract and Background
This PR was created while working on task RJD-1921.
Outline
- When Ego control mode has just changed to self-driving from manual driving ( , which is simulated with FollowTrajectoryAction ) ,
- Ego’s pose changed to the opposite direction
The fix reduced the issue frequency from ~1/30 to ~3/400 (jira comment as evidence).
Known Limitations
The issue with FollowTrajectoryAction for waypoints without timestamps still exists.
In this case, the entity can overshoot the final waypoint, failing the required conditions.
This may rotate the entity by 180° and cause it to "opposite direction " at the final waypoint.
This issue occurs ~3/400 in this scenario.
Work on this issue continues in task RJD-1921.