| Robotic systems have proven effective in many domains. Some robotic domains, such as mass casualty response, require close coupling between the humans and robots that are able to adapt to the environment and tasks for successful completion of a mission. The planning problem uses the task requirements and the team members' capabilities to determine a set of actions that will accomplish the task. However, planning problem difficulty increases exponentially with the number of expressive features included in the mission description, the number of agents, and the number of tasks. Each additional agent and task makes a problem more difficult and the problem can quickly become intractable, especially for missions requiring rapid decisions.;A natural first step to address planning problem difficulty is assigning agents (humans or robots) to teams based on each agent's synergistic capabilities and the requirements of each task, a problem known as coalition formation. However, coalition formation is also a difficult problem to solve and does not produce a plan for completing the task. Once coalition formation allocates agents to tasks, the question that remains is how the agents will complete their assigned tasks.;The research goal is to quickly find approximate solutions to the coalition formation problem and use the results to focus the planning problem on individual tasks with coalitions of the available agents. Unfortunately, there are no guarantees that planning will be able to derive a plan for each coalition-task allocation produced by coalition formation, resulting in a nonexecutable coalition. Planning tools must be resilient to nonexecutable coalitions; the tool must be able to use information from planning to intelligently augment the generated coalitions in order to transition nonexecutable coalitions to executable coalitions with a plan to accomplish their assigned task.;Planning for each coalition and task individually will greatly reduce problem difficulty at the cost of being unable to consider the interactions between the plans for each task. Conversely, planning for multiple tasks simultaneously allows the interactions of the tasks to be considered in the resulting plan, but increases problem difficulty. Intelligently merging select tasks and coalitions that are likely to interact can balance problem difficulty with the need to derive plans by considering possible agent and task interactions. |
