Search

A cost-effective architecture for the control of robot manipulators based on functional decomposition of the equations of motion is described. The Lagrange-Euler( LE) and the Newton-Euler( NE) formulations are used for decomposition. According to real-time control criterion, the LE equations are not suitable for implementation using currently available hardware because the required number of computations is too high, even after taking the inherent parallelism into account. However, the recursive nature of the Newton-Euler equations of motion lend themselves to being decomposed to terms used to generate the recursive forward and backward formulations. A special architecture implemented on a network of transputers is proposed which takes advantage of both the parallelism and seriallism of the NE equations and the ease of building communication channel provided by the transputers and Occam language. This proposed controller model can be best defined as a macro level pipeline. Based on this model, both floating point computation and fixed point computation results are presented for performance comparison.