Gesture based master-slave controller for a social humanoid robot

In this paper, a master-slave control system for a humanoid robot for social interaction tasks is proposed. The master-controller is a desktop anthropomorphic controller device that can gather set of joint angles for two arms (7 joints in each arm) in real-time when it is manipulated by an operator and the head motion command is acquired using a 9 axis IMU embedded in the hat worn by an operator. The master-controller is designed to be used with different humanoid robots. Thus, the master-slave control scheme with an automatic joint compensation is proposed. The decision-tree technique is employed in the gesture classification procedure. The joint commands that are sent to the slave humanoid robot are automatically compensated to ensure that both of its end-effectors reach the correct position for different gestures. The proposed controller was tested with different design parameters on the simulation of a humanoid robot.