Slope-Walking of a Biped Robot with K Nearest Neighbor method

Abstract

This paper proposes a walking pattern generation method with k nearest neighbors (K-NN) so that a bipedal walking robot can walk with compensation for environmental changes. We make up some walking path, and save them to a database, before the robot walks on slope. And we select a walking path for a suitable condition. When the robot walks on slope, we save the measured walking path to the database. Herewith, the robot has a learning function and can cope with environmental change. The proposed controller is examined and evaluated from simulations. The walking of the robot is simulated by using the proposed method, and its effectiveness is verified.