The task-level robot programming with computations of language processing, kinematics, dynamics, path planning, control, collision avoidance, and sensory fusion for multiple robot manipulators will be achieved in real time via coarse-grain vectorization and parallelization using an industrial standard VMEbus-based distributed computing system consisting of multiple scalar and vector processors. This computer system is assembled with off- the-shelf commercial VME boards. The results of the project can, therefore, be easily transferred for practical industrial applications and should thus make a meaningful and significant contribution towards the development of task-level multi-robot manipulation systems. Since the programming environment is C/Unix compatible, the robot programming can draw on the large existing body of skilled C programmers and Unix users. This research investigates task-level programming for manipulation of multiple robots, and studies its underlying principles and algorithms for force and motion coordinations, collision detection and collision avoidance, and sensor fusion. The research developes a theory for sensor-based multi-robot manipulation and force coordination with extensive information sharing among different robots, algorithms for collision detection and collision avoidance of moving objects, and techniques for force and vision sensor integration with robot motions. A set of standard task-level functions are developed. Through these task- level functions, a complicated robot task can be conveniently accomplished. The theory, algorithms, and task-level robot programming are validated through computer simulations as well as physical experimentations.
