Trajectory Generation for Path-Accurate Jerk-Limited Sensor-Based Path Corrections of Robot Arms

Kurzfassung

In previous work of the authors, trajectory generation for immediate path-accurate stopping is presented. It turns out that, with some extensions, this method is also applicable to generic trajectory generation, especially for computing a velocity profile for online modifications of the desired path, e.g. due to unexpected sensor data. The method computes a trajectory that satisfies constraints on the velocity, the acceleration, and the jerk, while staying as close as possible to the sensed desired path. A feasible solution is always found, even if exact path accuracy is not reachable. In contrast to time-optimal execution of the desired path, in this presentation the goal is to synchronize with the original robot program. So, e.g. after the execution of a sensed deviation the robot continues the original trajectory, maybe shifted by a sensed offset. The new trajectory is computed by iterative forward scaling and backtracking, where in contrast to other previous work the arc length is interpolated (ALI). In this way the path accuracy is superior to that with direct pose interpolation (DPI), which may feature undesired blending of subsequent path segments. Because of its time-efficiency the algorithm can be applied in each sampling step, e.g. every 4 ms for a standard KUKA robot with RSI interface.