Deformable trees - exploiting local obstacle avoidance

This work describes a hybrid trajectory planning approach for redundant robots, combining a local obstacle avoidance control framework with a global sampling-based planning component. The complexity induced by the high dimensionality of the configuration space is counteracted by shifting the global search to a low dimensional task space representation. It is shown how an advanced autonomy of the controller to locally circumvent obstacles can be exploited to speed up global planning and to increase the robustness of planned trajectories against small obstacle disturbances.