-
Notifications
You must be signed in to change notification settings - Fork 2
9th July, 2016
So, I encountered issues related to full body planning:
-
Inflation radius of base : This is the radius with which all the obstacles are inflated in the 2d costmap before doing a search. The admissible values is the inscribed radius of the robot base (as the base is square). However, while using an admissible value, quite often narrow passages are not feasible for the robot base. This gets it stuck in a local minimum and it comes out of it only after expanding all the arm positions. The way to avoid it is to have a larger (inadmissible) inflation radius. But this can lead to the planner failing if the goal base is close to the obstacle (i.e in the goal state, the base is feasible but since we have overestimated the base size, the planner fails).
I am currently trying to fix this by having two costmaps, one with admissible radius and another with an inadmissible radius. We will keep an admissible 2d search heuristic and use the inadmissible one everywhere else.