Risk evaluation of ground surface using multichannel foot sensors for biped robots

In this paper, a method to calculate a risk index for the walking of biped robots in uneven terrain using the information of foot sensors is proposed. First, a feature for detecting unexpected large disturbances is extracted from the foot sensor observations using the 4th order moment. Next, the extracted multichannel features are mapped into a single risk index using the logistic regression model. In the experiment, the proposed index showed a high value indicating risk before the robot fell and when an equivalent large force was observed.