Continuous CMAC-QRLS and Its Systolic Array

摘要

Conventionally, least mean square rule (LMS) is used to update the weights of cerebellar model articulation controller (CMAC). The algorithm of CMAC-RLS which applies recursive least square algorithm (RLS) to update the weights of CMAC has proved to be a good tool for modeling on line. Based on QR decomposition, a simplified algorithm of CMAC-RLS named CMAC-QRLS is brought forward next and its corresponding systolic array is also designed. Combining with B-splines, we further devise the systolic array of continuous CMAC-QRLS. The simulation results reveal the good performance of this proposed algorithm.