Human intention recognition for robot walking helper using ANFIS

Abstract

In this paper, upon a passive robot walking helper developed in our laboratory, referred to as i-go, we attach the force-sensing grips to its two handles, which are used to effectively recognize the intention of the user. We use the Lasso model for user intention recognition, so that we can infer the relationship between the user' intention and measured pushing-pulling force. We also use the PCA algorithm to obtain the weighting for each intention. With the intention weighting identified, we then propose a fuzzy controller to generate proper braking force for i-go. In addition, to adapt to different users, we apply the ANFIS to tune the parameters of the fuzzy controller to achieve more accurate assistance.