Research on sensing information forecasting for Power Assist Walking Legs

This paper presents a novel sensing information forecasting algorithm based on time series analysis for the power assist walking legs (PAWL). The goal of this algorithm is to improve the dynamic response of the exoskeleton. The algorithm is built up with the autoregressive (AR) model, the recursive least square (RLS) method and the final prediction error (FPE) criterion. The method of RLS is utilized to make the on-line parameters estimation, and the FPE criterion is used to select the order of AR model. The forecasting algorithm is designed to be used on-line and to make predictions of force sensing information to ensure the real-time quality of the whole system. The algorithm can be categorized into two types: one step forecasting method and multi-step forecasting method. Meanwhile, we make some simulations to verify the validity of this algorithm. At the end of this paper, the correlative experiments have been carried out, and the results demonstrate this sensing information forecasting algorithm can predict the value and the trend of the sensing signal precisely and effectively.