Robust Position/Force Control of Constrained Flexible Joint Robots with Constraint Uncertainties

Abstract

AbstractA novel robust control method for simultaneous position/force control of constrained flexible joint robots is proposed. The facts that the uncertainties make the usual control task unsolvable and that the equations of the controlled system are differential-algebraic make the problem dealt with considerably demanding. In order to overcome the unsolvability problem due to the constraint uncertainties the position control task is redefined in a practical way such that only a suitable subgroup of the link positions are driven to their desired trajectories. To determine the elements of the subgroup a simple algorithm of practical relevance is proposed. Under certain smoothness conditions to the contact surfaces, it is demonstrated that the position control problem can dynamically be isolated from the force control. Thus, it becomes possible to handle the position and force control tasks separately. The most significant advantage of the separation of the position and force control tasks is that it makes possible to adapt the position control methods known from free robots. Each joint is used in either position control or force control. The proposed position controller has a cascaded structure: First, trajectories for joint positions that drive the link positions to their desired values are calculated. Then, the joint torques that drive the joint positions to their calculated values are determined. A further significant benefit of the separation of the position and force control tasks arises in the force control such that the transformed equations are linear and any linear robust control approach can be used for the force control. The whole controller requires the measurement of the link and joint positions, the link and joint velocities and the contact forces, and allows modeling uncertainties in the equations of both the robot dynamics and the contact surfaces. The proposed control method is also confirmed by simulations.