Open carlosjoserg opened 9 years ago
Hi, I agree with @carlosjoserg. I think it can be resolved with configuration files. The only one considering minimization of a metric in T^n is the RRT*. I think it should be possible (not sure) to change the const function to one considering a metric either in SE(3) or a weighted one in T^n.
The problem didn't seem to be the planner, but rather the IK solver which got towards strange configurations. This should be solved, now everything is planned in joint space and IK low-level implementation supports looking for close configurations. Anyway, more testing might be needed before closing the issue
:+1:
The default configuration of all planners in MoveIt! does not consider the minimization of the joint path length in this situation, and the result is obtaining long trajectories to arrive to the same or very close Cartesian position of the end-effector.
@hamalMarino suggested moving to the MoveIit! interface and not using the MoveIt!-ROS interface @carlosjoserg thinks it might be resolved with configuration files.
In any case, this issue must be resolved between the low-level planner and the hardware teams.