Reflex Control

Abstract

Reflexes have been viewed as integrated motions with the centrally generated motors commands to produce adaptive movement. Lateral and frontal disturbances during locomotion due to terrain irregularities have been dealt with using conventional sensory feedback that was realized based on the inverse pendulum model. This chapter deals with reflexes that highly adapt and control the movement of the humanoid robot when a large disturbance occurs. The reflex action consists of modulating the motors’ commands by the outputs from both the force sensors located under the robot legs and the gyro sensor located at the robot’s upper body. A primitive neural network can deal with simple reflexes. These reflexes can be improved further to robustly address particular classes of sudden events. Eventually, in this chapter, primitive reflex against sudden obstacles is improved by the afferent signals in order to be more adaptable and robust against unexpected obstacle hitting the robot sole plate at random locations. The modified adaptive reflex consists of increasing the support polygon by controlling the ankle joint of the leg touching the obstacle. With such adaptation, the reflex response can be coordinated and modulated with locomotion controller’s outputs to achieve an intended stabilizing behavior of the robot.