Flexible Group Behavior: Lessons Learned Building Virtual Commanders

In the course of developing intelligent synthetic forces and commanders we have learned a number of lessons about what it takes to produce flexible group behavior: (1) simply putting together entities with intelligent, reactive behavior is not sufficient to produce team or group behaviors; (2) commanders need the ability to plan or project into the future; but, (3) planning alone is not sufficient. A commander also needs to be aware of the situation, monitor the plan’s execution, and re-plan, as necessary. (4) Even intelligent, continuous planning is not sufficient to produce flexible group behavior. The commander agent also needs a form of social intelligence that enables it to act as a part of an organization of other agents, which includes superiors, subordinates, and peers. (5) A Divide and Conquer (highly modular) approach to implementing an agent does not seem feasible. Situation awareness, planning, execution, monitoring, re-planning, and collaboration are highly interdependent. We addressed these issues by developing techniques that enable commanders to plan continuously and to exercise social intelligence.