2023 年 89 巻 924 号 p. 23-00104
Pneumatic rubber artificial muscles are nearly ideal for power assistance, medical care, and rehabilitation due to their lightweight drive unit made of rubber, which is soft and safe for human use. However, they have complex nonlinearities, including input saturation, hysteresis, and changes in static characteristics due to shrinkage. Additionally, when used with a human subject, the artificial muscle control system’s characteristics may vary based on the person’s physique, fitness, temperature, and the rubber’s deterioration. Therefore, it is desirable to be able to auto-tune the PID and other controllers in a short time before use. Data-driven controls such as FRIT and VRFT are suitable for auto-tuning in a short time. Among them, V-Tiger presented a procedure for predicting the actual closed-loop response and its Confidence interval. However, V-Tiger is not applicable to nonlinear systems with input saturation. This paper describes a procedure for (1) tuning PID gains by predicting the closed-loop response with V-Tiger in the presence of input saturation, (2) obtaining Confidence intervals for the predicted response affected by input saturation, and (3) how to use the Confidence intervals to apply this method to nonlinear systems.
