The result path may hit the obs

[kintzhao]

The result path may hit the obs.

test map file
test map file narrow_test-nav.zip

[LiJiangnanBit]

Maybe the reference line smoothing went wrong... you can display ref line on rviz to check it out. this part is not very stable, replace it with your own algorithm :)

[kintzhao]

I create the reference line by searched path with dijkstra, the opt_path is not better

[kintzhao]

I fork and modify the code for my diff-robot. kintzhao/path_optimizer_2

[cshang412]

i met the same problem for ackerman vehicle

[cshang412]

seems like the issue was caused after reference path smoother, from the image i posted before , the second stage optimization after reference path ignored the upper bound and lower bound which was caculated before. could you help us to fix this the final optimization stage ? thanks very much ! @LiJiangnanBit

[LiJiangnanBit]

seems like the issue was caused after reference path smoother, from the image i posted before , the second stage optimization after reference path ignored the upper bound and lower bound which was caculated before. could you help us to fix this the final optimization stage ? thanks very much ! @LiJiangnanBit

It seems impossible to stay within the drivable area with the init position and heading that you chose. In the optimization problem, the vehicle dynamics is a hard constraint so it's possible that the result violates the boundary constraint. :) @cshang412

[cshang412]

Thanks for your reply ! Yeah，you are totally right, but when i use this part in an dynamic Env, running at 10-20HZ, the reference point or the start_pose & dest-pose all changed very fast, this situation does happen a lot , It's very unsafe, that's why i choose an tricky corner case for start pose and dest-pose to test the algorithorm. i tried to enlarge the safe margin in the config file, not working very well also . still trying to figure out how to handle this situation. @LiJiangnanBit