A machine learning approach to falling detection and avoidance for biped robots

A falling avoidance of biped robots is an important research topic to use the robot in a human life environment. In this paper, we propose a machine learning approach to falling detection and avoidacne for biped robots. Support Vector Machine (SVM) is used as the machine learning algorithm and it detects the falling state of the robot based on acceleration value of torso and center of pressure value of the robot. When the falling is detected, the reaction module produces gait for extending areas of supporting polygon of the robot. The main contribution of the paper is falling detection of the biped robot based on the sensor data and machine learning algorithm without explicit dynamic parameters of the robot and predefined threshold value.