2014 年 50 巻 5 号 p. 411-417
This paper discusses about estimating nine parameters of a nonlinear model of a McKibben pneumatic artificial muscle (PAM) system in a game-theoretic learning way. An algorithm for the learning is constructed to find a Nash equilibrium in which the parameters induce transient and output responses close to corresponding experimental reference data. With a practical PAM system we made for validation, then, it is confirmed that the learning algorithm can estimate better parameter values in the sense of how close the computed responses and the reference data are and of how long it takes to acquire the parameter values.