This project is aimed at substantially increasing the flexibility, autonomy, and system robustness of heterogeneous multi-robot teams by changing the fundamental abstraction that is used to represent robot competences from the typical "task" abstraction to our proposed ?schema? abstraction. The inputs and outputs of these schemas are characterized in terms of their semantic information content, which our work exploits to automatically reconfigure the schemas to address the teaming task at hand. This new abstraction allows humans to design task-independent software that can be autonomously reconfigured by the robot team in response to dynamic changes in the environment, and the current task needs of the team. In doing this, we are able to simultaneously obtain a number of significant, wide-ranging new benefits in coalescent multi-robot teaming, including: (1) enabling robots to generate solutions to new tasks that were not explicitly programmed by the human designer, but instead consist of new, automated combinations of low-level building blocks, or schemas; (2) enabling robot team members to automatically generate coalescent task solutions based on sensor-sharing across team members, in configurations not previously explicitly defined by the human designer; and (3) enabling flexible software code reuse from one robot teaming application to another.
