This paper describes the implementation of an algorithm for the on-line optimal path control of a robotic manipulator using an array of Texas Instruments TMS9900 microprocessors and a global memory. In contrast to work by other investigators, this approach utilizes a horizontal decomposition which assigns a single processor to each joint for computing kinematics, but distributes the computation of control by coordinant axes. In the present configuration, the arm is modeled as a point mass under gravity loading and control is computed in Cartesian space utilizing minimum energy techniques.