Supervised Autonomous Interaction in Unknown Territories - A Concept for Industrial Applications in the Near Future

Despite recent advances in domestic and collaborative robotics, the main application scenarios for robots in industry are still well-defined manufacturing processes. The few existing solutions for more complex tasks, like bin picking or collaborative packing are either not flexible to faults and slight changes in processes or come along with a high integration effort. In our contribution, we propose an intelligent and flexible framework for operating robots in future industrial applications, where they work in unknown territories and share a common workspace with humans. The approach is based on autonomous operation augmented by on-demand human supervision and is evaluated within a bin picking scenario. We combine a motion planning module enabling real-time collision avoidance, a machine-learning based grasp pose estimator and supervised fault handling that computes and evaluates recovery strategies with different levels of autonomy. Preliminary experiments show that the projected system is able to perform online obstacle avoidance and to find feasible fault recovery strategies.