The complex dynamics of manipulators make design of dynamic control laws for them very difficult. Today’s industrial robots employ dynamic control laws which neglect the complex dynamics, thus leading to less than optimal response. The computational bottleneck of many advanced control schemes is an algorithm for computation of the actuator torques (forces) required to produce desired joint accelerations, for a given set of joint velocities and angles (displacements). The algorithm must be evaluated by the control computer about 100 times per second for the scheme to be effective. The previously known algorithms require large minicomputers to satisfy this speed requirement. The scheme described here, when implemented on a single processor, is about five times faster than the recursive Newton-Euler algorithm. When implemented on two processors, it is about ten times faster. The computational efficiency is achieved by partitioning the manipulator into two parts for modeling purposes. The modeling technique uses no approximations and can be applied to manipulators with up to seven links, which includes most designs presently in use.

This content is only available via PDF.
You do not currently have access to this content.