Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version


This is already possible, at least up to some extent. We have been using joint space planners for planning simultaneous motions of two 7-dof arms since the boxturtle release of ROS. Works like a charm. It's all in how you setup the different planning groups. Also, make sure that you select a planner that is compatible with multiple endpoints. We've had good results with lazy RRTs.

Our usual workflow calls directly OMPL and the trajectory filters, but back in the day I tested it with the move_arm action as well. We never got around to exposing an IK solver for task space goals, but I don't see why that shouldn't be possible, and it's in our TODO list.

One thing we have not (yet) incorporated in our motion planning is torso joints. Since these joints are located before the arms branch out, shortcut smoother filters don't work well. In such cases, using a planner with optimality guarantees, like RRT* would be the way to go. Also, using a weighted distance metric would allow to favor motions of more distal joints over proximal ones.

@Sachin: What would be the scope of the multi-arm pipeline?.