Random Re-Connection Leaning Algorithm of CMAC Model in Prosthetic Knee Control

A new learning algorithm of random re-connection (RRC) originating from the problem of mapping precision of CMAC controller used for prosthetic knee, which connects input layer with neural cell layer, is put forward from the view point of structure optimization in this paper. After the learning process of RRC, the in-degree distribution of neural cell becomes to follow power-law, which indicates that the effect of every cell in pattern identification is different. The simulation result of applying the RRC algorithm to prosthesis control shows that the algorithm of random re-connection can significantly improve the mapping precision of CMAC network model.