Control Architecures for Autonomous Teams (CAAT)

"Cooperative, autonomous control is required to realize the full potential of autonomous UAVs in a tactical military setting. This proposal addresses the cooperative platform control problem where the control task is to coordinate teams of small,expendable, platforms to perform Intelligence, Surveillance, and Reconnaissance (ISR) tasks. When viewed as a centralized control problem under uncertainty, Stochastic Dynamic Programming (SDP) provides the optimal control strategy for this stochasticproblem; however, this solution approach is not computationally feasible for most problems, and requires excessive centralization of information and computation. Moreover, the impact of imperfect communication performance decreases the overall performanceachievable by large, centrally controlled teams. We propose innovative distributed control architectures that extend previous ALPHATECH Approximate Dynamic Programming (ADP) successes to cooperative control of teams of UAVs. ADP is used as a basis forcontrol of centralized and decentralized architectures. The Phase I work, organized around a simulation-based trade study, begins with the evaluation of a centralized architecture. Insights into the centralized implementation are used to identify andaddress significant issues organic to all decentralized architectures. Phase II extends the work of Phase I by incorporating additional dimensions in the trade space and increasing the size and realis