This thesis presents a framework for guaranteeing safety in human-robot co-existence in a practical and industry-applicable way. We show how to control a robot such that its movement is verifiably safe with respect to user-defined criteria, while maintaining efficiency such that the robot does not stop unnecessarily, and is able to replan around obstacles.
To be able to guarantee safety online during operation, we develop fast methods to calculate the volume both the robot and the human may occupy during their respective trajectories, in a conservative way. In the case of the human, since the intention is unknown, we also study human movement, collecting data from a range of humans. From this, we develop prediction models which account even for unexpected movements, but do not overestimate the occupied space. Using these spatial occupancies to calculate trajectories in the shared workspace which are safe for the robot to execute, we guarantee safety in all human-robot co-working scenarios.
Finally, the approach is tested on humans. Test subjects react well to this approach in terms of trust, and efficiency of the robot is higher than a comparable approach from the state of the art. Also, as subjects work more with the robot, they adjust to its behaviour, reporting better understanding of the robot and feeling safer after experience compared to at first sight.
«
This thesis presents a framework for guaranteeing safety in human-robot co-existence in a practical and industry-applicable way. We show how to control a robot such that its movement is verifiably safe with respect to user-defined criteria, while maintaining efficiency such that the robot does not stop unnecessarily, and is able to replan around obstacles.
To be able to guarantee safety online during operation, we develop fast methods to calculate the volume both the robot and the human may occ...
»