Biological and Cognitive Foundations of Holistic Information Fusion

The study will formulate a novel biologically inspired approach to the development of general-purpose, holistic information fusion algorithms and architecture. This approach avoids fundamental limitations of conventional data fusion methods caused by ad hoc partitioning of the problem domain into compartmentalized subprocesses. As a result, conventional methods do not scaleup, facing combinatorial explosion as the size of the domain grows. In the proposed system the domain is represented as a unified integral structure capable of adaptive self-partitioning depending on the changing domain conditions and user-specified goals and constraints. The system is inherently robust, allowing near-optimal goal-satisfaction decisions under noise-corrupted domain inputs. The system's architecture and operation principles are derived from the recent models of human nervous system and processes underlying robustness and efficiency of human performance.