Fast computation algorithm for robot dynamics and its implementation

The authors propose an architecture of parallel processing to increase the computational speed of the matrix/vector operation for robot dynamics. The system is formed by nine processor elements with X-bus and Y-bus. This array processor is able to achieve zero data transfer time by using data sharing on the bus. The multiplication of 3*3 rotation matrices requires only three multiplications and two additions. The vector inner products and vector additions can be performed in one multiplication and two additions (or one addition), respectively. It is shown that the proposed architecture is more economic and more efficient than other existing approaches. Furthermore, the above architecture has been implemented on three IBM-PC boards by using the Intel 8088 CPU.<
>