S Ingle - Operator Control of Multi - Robot Teams

There is growing recognition that many applications of robots will require a human operator to supervise and control multiple robots that collaborate to achieve the operator’s goals. However, the bulk of existing work in this area assumes that robots are independent of each other, and thus ignores key challenges and opportunities in monitoring and operating tightly-coordinating teams. This thesis takes steps to address these open issues. First, we address the challenge of effectively monitoring multiple coordinating robots. We introduce a graphical socially-attentive display that explicitly shows the state of coordination in the team, in terms of the robots’ state with respect to each other. As a result, the operator can easily detect coordination failures, even before these cause overall failure in the task. Second, we show that in resolving contingencies (callrequests), an opportunity exists for taking advantage of the robots’ teamwork, to allow the robots to actively assist the operator. We propose a distributed approach to call-request resolution (including two variations), and an implementation method for behavior-based robots. This implementation method allows the operator to quickly switch control between robots, even while they are active. We evaluate all of these techniques in several multi-robot tasks, in experiments with up to 25 operators, each controlling multiple robots. The results show significant quantitative and qualitative improvements in task completion times, number of coordination failures, and performance consistency across operators.