We present a general framework for planning of gross motions for dual-arm robots. The environment is not known in advance but is assumed to be static during the execution of each motion. Tasks are repeatedly assigned to the robot although only a few tasks into the future may be known at the time. If the environment changes, the robot must be alerted, and the motions for the tasks ahead must be recomputed. As an application example, a pair of robots may cooperate on emptying a bin, one object at a time. The environment changes as the objects are removed from the bin, and the grasping of an object itself changes the set of feasible motions. New tasks are assigned to the robot as new objects become visible in the bin.
展开▼