Online near time-optimal trajectory planning for industrial robots

A novel trajectory planning approach is presented suitable for online industrial robot applications along short path segments such as spot-welding. The proposed method generates trajectories that are computationally efficient, dynamically near time-optimal, and maintain path-following integrity in high-frequency planning-and-control cycles. The method is based on the well-known path constrained time-optimal motion. We show that this trajectory can be quickly approximated with trapezoidal velocity profiles, resulting in near time-optimal trajectories, requiring only four robot dynamics computations per path segment. For continuous motions, a method to safely transit between adjacent optimal path segments within geometric bounds is also presented. We then show how the generated second order trajectory can be successfully used with a generic control loop by adopting feedforward control based on an elastic model. A real-world experiment with a 6DOF industrial robot validates our approach.