A Study of Effectiveness of Dynamically Varying Sampling Rate for State-action Pair Prediction

When considering an action decision based on a future prediction, it is necessary to know the property of a disturbance signal from the outside environment. On the other hand, the acquisition of the property of the disturbance signal depends on the specifications of the target processor, in particular, its sensor resolution or processing ability. Therefore, sampling rate settings are limited by hardware specifications. In contrast, future prediction using machine learning is based on the tendency obtained through past training or learning. In this type of situation, the learning time increases in proportional to the amount of training data. In this paper, we focus on reducing the amount of training data. In particular, we consider the situation where a periodic disturbance signal occurs. From this perspective, we propose a method that adjusts the sampling rate dynamically based on the frequency of the disturbance. The results of verification experiments confirm that the proposed method can adjust the sampling rate according to the frequency property of the disturbance signal. Moreover, it is confirmed that each control response corresponds better to the disturbance than in the case where a fixed sampling rate is applied.