Dynamic Control And Formal Models Of Multi Agent Interactions And Behaviors

New Multi-Agent System (MAS) approaches to complex DoD problems hold the promise of previously unrealized levels of autonomy, adaptability, and flexibility of agent-controlled systems. These systems will provide essential capabilities in command and control, surveillance, automated targeting and weapons delivery, and biochem monitoring. ALPHATECH's work has been focused on three areas. First is the development of an Open Experimentation Framework to facilitate research, evaluation, and characterization of the emerging science of Multi-Agent Systems. Second is the design and development of the Testbed for Taskable Agent Systems (TTAS), which is a software environment facilitating experimentation with disparate agent technologies and evaluation of critical design elements of Multi-Agent Systems. Lastly, our theoretical research developing cooperative methods for machine learning in Multi-Agent Systems and designing goal-directed agents that make and adapt decisions in a heterogeneous dynamic environment within a coherent mathematical framework of dynamic programming and Partially Observable Markov Decision Processes.