A CMAC with Content Addressable Memory

Abstract

A new design scheme of Cerebellar Model Articular Controller (CMAC) is presented in this article. The controller is designed with a content addressable memory (CAM) to replace the function of hash-coding method, which consumes more computation efforts, in the traditional CMAC for memory space consideration. Therefore, the address mapping of the proposed CMAC is comparatively different from the hash-coding method. The CAM, which is capable of fast comparison, can immediately determine in parallel if there is any akin data in memory as the input data presents. If a match doesn’t occur after comparison, the activated address is then stored in a vacant cell of the CAM indexed by a circular incremental pointer. The memory collision should be able to be avoided unless the memory is all occupied. Thereby, memory utilization rate can be improved to 100%. Meanwhile, control noise can be suppressed significantly. In content addressing, the associated mask function, which is triggered while the CAM is full, can decrease the probability of collision and improve control performance. Simulation results of function approximation and truck backer-upper control indicate the proposed CMAC is explicitly superior to the conventional CMACs