Parallel computational methods based on symbolic robot models - theory and application on pipelined processor architectures

摘要

Increasing the speed of inverse dynamics computation is essential for improving the characteristics of robot control systems. This is achieved by reducing the numerical complexity of the models and by introducing parallelism in model computation. Customized symbolic models with near minimum numerical complexity are used as a basis for examination of parallelism in inverse dynamic robot models based on Newton-Euler's equations. Furthermore, a linear array of pipelined processors as well as an array processor architecture is used for demonstrating proposed scheduling algorithms.