FUJITA Tomohiro
Department / Course College of Science and Engineering Department of Electronic and Computer Engineering
Title / Position Professor
Date 2020/09/25
Presentation Theme Cerebellar Model Articulation Controller with Threshold Learning Cell
Conference 2020 59th Annual Conference of the Society of Instrument and Control Engineers of Japan (SICE)
Promoters the Society of Instrument and Control Engineers (SICE)
Location オンライン
Conference Type International
Presentation Type Speech (General)
Contribution Type Collaborative
Holding period 2020/09/23 ~ 2020/09/26
Publisher and common publisher Yamada Soichiro, Kumaki Takeshi, Ogura Takeshi, Fujita Tomohiro
Details We propose a cerebellar model articulation controller (CMAC) with a threshold learning cell (TLC). The CMAC is widely used for neural-network-based control, and we introduce the TLC to the CMAC. This CMAC + TLC has a simple structure and its learning algorithm consists only of comparison, increment, decrement, and count. Thanks to the simplicity of this learning algorithm, the advantage of the CMAC + TLC is its easy hardware implementation. We experimentally evaluated its function-approximation accuracy. The results indicate that the CMAC + TLC has a better approximation ability than the original CMAC.