A Dataflow Multiprocessor System for Robot Arm Control

With the development of robot control theory, there has been an increased demand for more efficient robot control schemes. The real-time computation of applied torques in nonlinear and strongly coupled multi-link systems is compli cated and time consuming. This paper proposes a dataflow multiprocessor system for robot arm control. The proposed architecture computes the applied torques of a multi-link system using the Newton-Euler state space formulation. Sim ulation studies have verified the effectiveness of the proposed architecture by producing execution times that are fast enough for real-time control application. The architecture that explores the maximum parallelism would consist of 1834 processing elements; however, a reasonable engineering solution consists of 42 processing elements. If this dataflow architecture with reduced number of processing elements is implemented using currently available fast-processing ele ments (40 ns multipliers and adders), it calculates the torque for a six-link robotic system in 1.02 ms. The system is simu lated using Transputers and the language Occam.