search

rayvburn / humap_local_planner

Human-aware robot trajectory planner using a hybrid trajectory candidates generation and spatiotemporal cost functions
BSD 3-Clause "New" or "Revised" License
0 stars 1 forks source link

investigate causes of massive decelerations in cluttered areas of the world #60

Closed rayvburn closed 1 year ago

rayvburn commented 2 years ago

First, check if increasing internal force factor will do any good.

rayvburn commented 2 years ago

Everything seems that non-linear transformations of raw force vectors in SFM cause such issue. Loosening max_force_factor helped so far.

rayvburn commented 1 year ago

Seems that with the fallback trajectory generator (SimpleTrajectoryGenerator) this problem is somehow hidden by the emergent behaviour of the planner.

rayvburn commented 1 year ago

ATM not sure whether this issue was related to simulation performance degradation or robot's deceleration. If it is related to robot's deceleration it may happen that control loop takes too long (tested robot (TIAGo) has velocity command timeout of 500ms). The performance of simulation has not been observed recently.

rayvburn commented 1 year ago

Apparently, the Gazebo simulation slows down with any planner in some areas. It is rather unclear why it happens, but since other planners also suffer that issue, #60 won't be investigated in the simulation performance aspects.

rayvburn commented 1 year ago

It has not been noticable for some time. It could have been caused by the usage of Eigen types. See this https://github.com/rayvburn/hubero_local_planner/pull/54/commits/7e8a24774ca6e11cba536644395aec5f15f38f23 and that commit in a dependent package social_nav_utils https://github.com/rayvburn/social_nav_utils/pull/6/commits/5e3e73a0d571770273244158b34276b983d172af