Online nonlinear reference shaping with end-point position feedback for human-like smooth reaching motion

A nonlinear reference shaping method for manipulators which are operated in living environments is proposed. It generates an intermediate reference position, and by connecting the position and the manipulator´s endpoint with virtual spring and damper, the robot acceleration is moderated and smooth reaching motion is realized. By feeding back the hand position of the manipulator to the proposed reference shaping controller, the motion of the robot becomes pliant against external forces applied to it without generating excessive pushing forces and large acceleration. Furthermore, according to the initial/desired hand position, the parameter of the reference shaping controller is changed dynamically to avoid large acceleration and generate smoother, human-like reaching motions. The convergence performance to a target hand position is never decreased with the proposed method. The validity of the proposed method is shown through computer simulations on a planar 4-DOF manipulator.