Fast Parallel Computation on Inverse Dynamics Equations of Articulated Manipulators by a Modified Task Division

A method for a task division which is needed to realize an efficient parallel processing on the inverse dynamics problem is proposed. In the preprocessing step of scheduling, calculating equations are divided into many tasks which become processing units. On dividing the given equations into a task set, if the task granularity is fine, the number of tasks which can be processed simultaneously tends to increase, whereas communication overhead between processors tends to increase. The purpose of this study is to develop a task dividing method for efficient implementation of parallel processing. The inverse dynamics equations are divided into tasks by equation level. Divided tasks are composed or divided additionally considering the precedence constraints inherent in the inverse dynamics problem.