Unsupervised real-time classification of cycle stages in collaborative robot applications

Recent robot applications often operate at a level of autonomy that does not need strictly structured environment within the robot cells. Human-Robot collaboration introduces further uncertain aspects, which in turn makes the production cycle non-deterministic. It renders difficulties for manufacturing companies regarding planning, control, quality assurance, etc. The latest robot controllers allow access to internal states (joint angles, accelerations, drive torques, etc.) at a high sampling rate, which can be acquired, processed by an appropriate service connected to the ERP system, developing cyber-physical control and monitoring infrastructure. This study introduces an online technique for automatic and unsupervised clustering of elementary steps in collaborative robot applications based on real-time force/torque data stream. The proposed unsupervised clustering solution utilizes dynamically trained one class support vector machines (OCSVM) to discover the states of the repetitive process. The method is demonstrated and evaluated in a realistic experimental environment. Initial results show promising performance characteristics.