Modular Affective Reasoning-based Versatile Introspective Architecture (MARVIN)

Robotic vehicles, unmanned systems, and critical command and control networks rely on systems that need to operate effectively in complex, dynamic, and adversarial environments. Currently fielded systems are not up to par—they can be immobilized by simple component failures. Future operations will not tolerate such brittleness, especially given the interconnectivity demanded by future concepts of network-centric warfare. Future fielded systems throughout the military will need more dynamic processing capabilities that can adaptively use resources. Self-Aware Computing attempts to resolve this problem through metacognitive approaches to generating dynamic capabilities based on research into human adaptation. However, these approaches tend to ignore the impact of human emotions in that adaptation. Emotions provide humans with approaches for instantly and broadly adapting to threats, for persisting adaptations across situations, for learning future reactions to similar events, and for effectively communicating those threats to others. We therefore propose to develop the Modular Affective Reasoning-Based Versatile Introspective Architecture (MARVIN) to develop components that enhance self-aware systems by exploiting human affective mechanisms to (1) adapt quickly to time-critical, situations, (2) learn to adapt quickly to new situations and carry learning over to additional situations, and (3) interact with other networked systems in an efficient social manner.