Real-time processing in robot dynamic control is difficult because of computational burden of robot dynamics algorithm. In this thesis, a parallel robot dynamics algorithm is presented and a parallel processing computer for robot dynamic control is implemented using three transputers. Recently, using the concepts of augmented body" and "barycenter", the robot dynamics algorithm has been developed. A merit of this algorithm is to increase the number of off-line calculations and decrease the number of on-line calculations. Therefore by reconstructing that robot dynamics algorithm, a parallelized robot dynamics algorithm to fit for characteristics and architecture of the parallel processing computer is proposed. Since generally communication between processors is slow in processing, the reduction of communication burden is emphasized in parallelizing of the robot dynamics algorithm. The transputer is a special purpose CPU for parallel processing. Transputers are used in networks to build a high performance concurrent system. A network of transputers and peripheral controllers is constructed using point-to-point communication. To gain most benefit from the transputer architecture, the whole system is programmed in Occam which is a high level language for concurrent applications. In executing robot dynamic control, it takes about 2500 microseconds to compute robot dynamics with uniprocessor, but we could reduce the computational time to 900 microseconds for the parallelized robot dynamics algorithm with multiprocessors.