This research area focuses on contextual abstractions to learn perception-action-communication loops for distributed intelligence, and collaborative, distributed inferencing and decision making algorithms. We will develop learning algorithms that are modular and composable, and algorithms that can adapt to available resources to enable scalability to large heterogeneous groups with human actors.
Heterogeneous Group Control
This research area addresses modeling of heterogeneous teams of humans, robots, sensors, tactical supercomputers, and tactical clouds that can be formed based on the context and the task. We will develop models of interactions between humans and other agents, and pursue formal approaches to composition of controllers, estimators and planners to design architectures and synthesize group behaviors.
Adaptive and Resilient Behaviors
This research area will enable robustness and adaptation of perception, inference, and machine learning modalities to changing conditions and adversarial inputs, and adaptive behaviors that mitigate growing uncertainty in the heterogeneous models. Ultimately, constrained by the limited resources at the individual level, we will develop macro-scale cooperation and situational awareness in optempo missions with resilience to agent failures, network disruption, data loss, and/or compromised communications.
Cross-disciplinary experimentation is an integral part of the alliance’s research program to explore and discover the interdependencies across the research areas. Experiments might incorporate autonomous physical agents, networked human experts, real-time distributed high performance computing, collection of large data sets to be shared across research areas and other elements. As the research matures it is desired to carry out complex large scale heterogeneous experimentation with autonomous agents, which may require indoor/outdoor sites.