Direct trajectory generation for vision-based obstacle avoidance

In this paper, a method for direct trajectory generation for robotic manipulators is proposed. The method is specifically designed for obstacle avoidance and can incorporate kinematic constraints into the avoidance motion. In particular, a point-to-point and multi-point trajectory generator is proposed in which different levels of constraints can be established on via-and end-points. The approach is direct in the sense that at every time instant a new trajectory is generated, which allows the trajectory to be changed at runtime. Moreover, when the final conditions of the trajectory are altered such that constraints would be violated, an optimization procedure is employed in order to extend the execution time of the trajectory. This effectively makes the trajectory time-optimal. The approach is experimentally verified with a 7-DOF anthropomorphic manipulator with time-synchronization of multiple trajectories.