Almost all industrial robots exhibit joint flexibility due to mechanical compliance of their gear boxes. In this paper we outline a design of an adaptive controller for flexible joint robots based on the arms energy. The desired actuator trajectory in a flexible joint robot is dependent not only on the desired kinematic trajectory of the link but also on the link dynamics. Unfortunately, link dynamic parameters are unknown in most cases, as a result the desired actuator trajectory is also unknown. To overcome this difficulty, a number of control schemes have suggested the use of acceleration and link jerk feedback. In this paper we describe a control scheme which does not use link jerk or acceleration. The control law we derive is based on the energy of the arm deviating from the desired trajectory and it has two stages with two corresponding adaptation laws. The first stage drives the actuator and the joints to a desired manifold, the second controller then seeks to drive the joints to their desired trajectory. On application of our first controller there is an apparent structural reduction of the order of the system. This apparent reduction in the structure is exploited by our second stage controller. Our control scheme does not require link acceleration or jerk measurements, and the numerical differentiation of the velocity signal, or the inversion of the inertial matrices are also unnecessary. Simulations are presented to verify the validity of the control scheme. The superiority of the proposed scheme over existing rigid robot adaptive schemes is also illustrated through simulation.
Date of this Version