Efficient computation of system-specific motion commands for serial and parallel robots based on differential flatness

This paper presents a universal method to reduce vibration and contouring errors of complex, nonlinear robotic systems during dynamic motions. The proposed method is based on differential flatness and is valid for serial and parallel robots. System-specific trajectories for motor position, velocity, and torque are generated online with minimized computational effort. Hereby, compliance and friction of the drive trains as well as automatically generated dynamics models are considered. Furthermore, a discrete approach is given to consider damping. The generated motion commands significantly relief the feedback control loop and, therewith, improve the overall system's motion behavior. To ensure time optimality, the trajectories are based on trapezoidal motion profiles, exploiting given constraints for velocity, acceleration, and jerk. The performance of the method as well as its sensitivity with respect to the quality of model parameters is studied and verified using an exemplary 3RRR parallel manipulator. It is shown, that the proposed method is able to outperform conventional computed torque control.