Envelope of human cognition for battlefield information processing agents

In this paper, we describe Envelope, an adaptive, symbolic, and hybrid cognitive architecture. Envelope integrates reactive and skill-based human behavior, which relies primarily on working memory, with conscious and knowledge-based problem-solving behavior, which relies heavily on long-term memory. Envelope relaxes various lower-level psychological constraints and becomes logically omniscient in its inference-based implementation. On the other hand, Envelope imposes somewhat high-level constraints on some aspects of intelligence in its use of hierarchical task and goal decompositions. This difference makes Envelope especially suitable for producing intelligent agents for a variety of goal-oriented information processing tasks involving large volumes of data. The specific implementation framework described here includes a high-level knowledge representation language for specifying tasks, goals, and knowledge, and various logical and probabilistic techniques for realizing the desired behavior of the architectural components. The generic nature of Envelope ensures its applicability across a variety of application domains. We describe our experience with Envelope and its potential use in various battlefield information processing tasks, including data fusion at various levels, decision-making, and planning for generating courses of action.