Revision history [back]

So, from the video, it looks like mojo never actually achieves its goal position to perform an in-place rotation. Likely this is because you've got the sim_time parameter set to 5 seconds with a minimum velocity of 0.1 meters/second. Trajectory rollout scores trajectories based on their endpoints which means that any trajectory with positive x velocity will move at least 0.5 meters forward. So, when the robot gets close to the goal, in-place rotations and trajectories that take long arcs towards the goal will score higher than the straight line path that goes too far past the goal. I'd suggest turning sim_time down to something like 1.0 or 1.5 seconds. Also, this will likely mean you can turn your path_distance_bias parameter down a bit and still get reasonable behavior. If you have the CPU, I might also up the vx_samples since you're using trajectory rollout rather than dwa and will benefit from additional sampling in the x velocity space, perhaps something in the 8-10 range. Hope this helps.