We are being called upon to protect our national security interests in progressively more complex and hostile environments. Major threats arise from asymmetric threats such as terrorism, guerilla attack, and other unconventional methods of warfare. The technology challenge for dealing with these asymmetric and extremely rapidly adapting adversaries in the battlefield are many, and of course, the battlefield itself is in a wide variety of terrains, in urban environments and in some cases also the homeland. These in turn require that we develop adaptive, intelligent, multi-agent cooperative control technologies over reliable, robust, and fault tolerant complex systems, with the capability of interacting with hard real-time constraints and the ability to reconfigure after failure.
On the program, we design and evaluate the adaptive hierarchical control of mixed autonomous and human operated semi-autonomous teams that deliver high levels of mission reliability despite uncertainty arising from rapidly evolving environments and malicious interference from an intelligent adversary. The design of architectures combining both hierarchical and heterarchical elements, the analytical foundations of interacting hybrid systems, the design of controllers for such systems that are robust against uncertainty, the management of rich sensory information from networked sensors among distributed and mobile teams; and the incorporation of human intervention in a mixed-initiative system are all key areas of our work. Our approach builds on the following research thrusts:
> Architecture design and analysis for dynamic,
adaptive planning
> Integration of rich multi-sensor information into virtual environments for incorporating human intervention in mission planning and execution
> Handling uncertainty and adversarial intent in
adaptive planning
In order to motivate the research agenda, we are developing three different scenarios that involve teams of autonomous and semi-autonomous multivehicle unmanned air vehicles (UAV) and unmanned ground vehicles. These scenarios will illustrate our vision and define experiments, technical demonstrations, and milestones over the course of the project. In each case, it is important to note that we will be planning in the face of an unknown environment and a hostile and intelligent adversary.
The three scenarios are: reconnaissance and intelligence (a robotic ranger force); mixed initiative engagement; and, recognition and tracking of unfriendlies. We will integrate rich multi-sensor information over an unreliable network by developing new classes of algorithms combining our recent work in omni-directional vision, the extraction of graphical models from video sequences, and the joint rendering of simulated (synthetic) environments with multi-sensor (real) data. The primary function of robotic UAVs and UGVs is to operate autonomously on specific tasks until a requested intervention arrives. Assessments of the effectiveness of our methods will be performed using cognitive models of the decision making process as well as real-time performance in experimental games. A key mathematical framework for the modeling of adversarial actions comes from the theory of games, and partially observable Markov decision processes. An engagement can be preceded by a learning phase when a number of scout UAVs and UGVs are sent out to prove and learn about an adversary's reactions for use in an engagement, using new graphical learning techniques.