Real-time multi-robot path planner based on a heuristic approach

The authors describe a real-time approach to solve the trajectory planning problem using the latest feedback of robot joint locations from robot controllers. The approach incorporates an efficient collision avoidance strategy to allow decisions to be made during execution by means of cylindrical robot link approximation. The developed algorithm has been verified with a dual-robot planning and control system for the mechanical assembly of a dish-washer power unit. The system consists of an ADEPTI and a PUMA 560 industrial robot running under the control of a Sun-4 Sparc 2 workstation at the high control level. A coarse motion planner is available to prevent the arm from colliding with stationary objects. In this dual-arm system, a task level multi-agent plan is generated to specify the logical sequence of assembly. From the execution results, it was found that even when the robot speed was varied from 25% to 80% of its full capacity, a collision-free path was found for each robot.<
>