A mixed-initiative framework for multi-agent human-robot teams

Robots and autonomous systems are expanding into more complex, dynamic, and human-centered operating domains. In such domains, neither full autonomy nor full human control are ideal and the appropriate level of autonomy is contextually-dependent. Mixed-initiative systems enable flexible autonomy, making them ideal for complex operating environments. However, most mixed-initiative research does not consider multi-agent human-robot systems, and a multi-agent mixed-initiative robot control framework is not readily available. This project explores the requirements for a mixed-initiative human-robot control framework and proposes an initial design. The mixed-initiative framework was deployed onto a mobile robot, where three different agents could take, transfer, share, enable, or disable robot control. Compared to single-agent systems, the multi-agent mixed-initiative system completed a series of navigation tasks more quickly. The multi-agent system also completed navigation tasks that a single-agent autonomous system could not. Further design improvements and applications for mixed-initiative systems are discussed.Mechanical Engineerin