Industrial Human-Robot Collaboration: Maximizing Performance While Maintaining Safety

For many years, separated autonomous robotic systems have been an essential component in industrial manufacturing. In particular, these heavy-payload robots perform a wide range of tasks, where high precision and repeatability is crucial. A flexible adaptation of fast changing tasks or environments as well as the interaction with humans can rather not be realized by these types of robots. Recently, a paradigm shift regarding customer demand could be observed. Short product life-cycles as well as increasing individualization of products require flexible manufacturing processes. Therefore, novel light-weight robot technology was developed, which enables the collaboration of humans and robots. In particular, highly productive robots are combined with the high flexibility of humans. However, only a few collaborative applications have been established in industry, which is mainly due to the low efficiency, i.e., large cycle times caused by safety regulations. The goal of this thesis is to maximize performance in collaborative applications, while maintaining safety. For this, assembly workplaces are analyzed, typical tasks identified, and the potential of collaborative robots is elaborated. Current safety regulations are analyzed in order to identify the challenges in safe human-robot collaboration. Then, a novel control method is presented, which enables intuitive, safe, and efficient control of robots. The Mirroring Human Arm Motions approach presents a velocity-limited trajectory generation, in particular, for orientations in quaternion space. This method is extended to an online via-point trajectory generation in order to enable an adjustment of velocity limits for guaranteeing safety in realtime. Furthermore, in collaborative applications particularly collisions with the human arm are likely to occur. Therefore, human-arm performance is analyzed and experiments similar to typical collaborative scenarios are executed, to determine the dynamic properties. By exploiting the obtained information on human arm dynamics, a novel approach to improve the performance of robot motions is presented. From the experiments, a simplified human arm model is derived, which enables the calculation of movements of the human into the path of the robot. With this approach, a maximum robot velocity depending on kinematic limitations of robots and human-in-the-loop constraints can be determined. This idea is further developed into a nonlinear optimization problem, where minimal-time motions are found and applications with low-cycle times can be realized. In order to enable flexible robot motions within the entire workspace of the robot, a generalization method using Dynamic Movement Primitives is presented. It contains a novel real-time consideration of spacial and kinematic constraints, to fulfill the requirements on safe human-robot collaboration. Experiments on a collaborative workbench prove the effectiveness of the presented methods. Finally, a novel airbag technology is proposed, which enables a protective coverage of dangerous tools and objects and protects humans against injuries, caused by a collision with the robot. The so called Robotic Airbag is inflated with pressured air to create a cushion around sharp edges of tool and object. Intrinsic safety is guaranteed, as the airbag is always inflated before initiating a robot motion. In order to exclude an affect of the tool functionality, the Robotic Airbag can be deflated whenever required. Experiments with a crash-test dummy, and finally with a volunteer, prove the functionality and compliance with current safety standards. In Summary, the presented methods in this thesis enable a significant improvement of efficiency and safety in collaborative applications