Passive computed torque algorithms for robots

Computed torque algorithms are used to compensate for the changing dynamics of robot manipulators in order to ensure that a constant level of damping is maintained for all configurations. There are three significant problems with existing computed torque algorithms. First, they are nonpassive and, as shown, can lead to unstable behavior; second, they make poor use of actuator capability; and third, they cannot be used to maintain a constant end-effector stiffness for force control tasks. A new class of passive computed torque controller which have guaranteed stability properties, utilize actuators effectively, and maintain constant end-effector stiffness is introduced. Network models are used to motivate and illustrate the approach.<>