An Approach to Supervisory Control of Multi-Robot Teams in Dynamic Domains

This paper explores an approach to human/multi-robot team interaction where a human provides supervisory instruction to a group of robots by assigning tasks and the robot team coordinates to execute the tasks autonomously. A novel, human-centric graph-based model is presented which captures the complexity of task scheduling problems in a dynamic setting and takes into account the spatial distribution of the locations of the tasks and the robots that can complete them. The focus is on problem domains which involve inter-dependent and multi-robot tasks requiring tightly-coupled coordination, occurring in dynamic environments where additional tasks may arrive over time. A user study was conducted to assess the efficacy of this graph-based model. Key factors have been identified, derived from the model, which impact how the human supervisors make task-assignment decisions. The findings presented here illustrate how these key factors capture the complexity of the task-assignment situation and correlate to the mental workload as reported by human supervisors.