Parallelism in manipulator dynamics

This paper presents two highly parallel "mathematically exact" formulations of the inverse Dynamics especially suited to high-speed implementations using VLSI devices. The first is a parallel formulation of the existing linear recursive Newton-Euler algorithm, with the real-time coefficients reduced almost two orders of magnitude. The second reports a new parallel algorithm which improves the linear time dependency to O(log2) of the number of joints, with small coefficients. Either formulation is susceptible to a systolic pipelined architecture, requiring four floating-point operations per complete set of joint torques. Hardware requirements are not excessive, and a VLSI implementation architecture is possible.