Design of Network Dynamics for Strategic Team-Competition

 Our research addresses the competitive-strategic domain for two opposing teams of agents, which we model in the general framework of a repeated game over network. We are motivated by applications in which, somewhat abstractly, a large collection of autonomous agents must play a game against an adversary or a team of adversaries. We note these applications need not involve explicit adversarial interactions. One example is the problem of controlling a large wind farm composed of numerous turbines: each rotating blade creates a downstream wake and every turbine faces the problem of setting rotation speed in the face of these complex aerodynamic interactions. Faster rotation by one turbine may lead to lower wind speed for nearby turbines, and further complications result from unpredictable changes in wind speed and direction. We would like to extract maximum power from the wind turbines; this comprises a "minimum regret game" as the turbine system should be congured to achieve as close to the maximum level of power output as could be achieved with perfect knowledge.