A key computational bottle-neck in real-time robot control is the calculation of the inverse kinematics which produces the time configuration history of the robot joints necessary to produce the desired end effector motion. The controller must solve the inverse kinematics problem to produce the correct trajectories. There are, in general, several reasons which complicate the solution even for nonredundant robot arms. Redundant arms have more joints than necessary to perform a certain motion of the end effector. In recent years some progress has been made in the design of custom VLSI coordinate rotation algorithms architectures suitable for redundant manipulator kinematics. Here the authors describe in detail their first results on the design and implementation of a high performance VLSI array specifically tailored for a robot application and the combination of the array with a high performance general purpose processor. The new architecture for efficiently computing the kinematic equations (direct and inverse) is valid for nonredundant as well as for redundant arms, and can use any type of high performance digital signal processor chip. The paper is well written and should be of interest for researchers and engineers from both mechanical and electrical engineering.