Looks like TEB handles differently (wrong?) line footprint models, compared with circle. Here I use a line model with both points at 0, 0, so it's effectively a point model:
You can see how weirdly planner works, alternating between being away from one obstacles on one side or the other.
However, an equivalent circle model (note how the solid the footprint marker is of the same size as on previous video), same radius, behaves as expected:
Looks like TEB handles differently (wrong?) line footprint models, compared with circle. Here I use a line model with both points at 0, 0, so it's effectively a point model:
https://user-images.githubusercontent.com/322610/145192287-5a900f0a-45c9-47a4-b5fe-001487db4bf4.mp4
You can see how weirdly planner works, alternating between being away from one obstacles on one side or the other.
However, an equivalent circle model (note how the solid the footprint marker is of the same size as on previous video), same radius, behaves as expected:
https://user-images.githubusercontent.com/322610/145192468-2e6362d9-5804-465b-a41d-04f374a66e3d.mp4