Director: Saibal Mukhopadhyay
Assistant Director: James Buckwalter
The vision for COGNISENSE is to design sensors that dynamically adapt to “what is being sensed” and “how sensed signals are processed” according to the real-time changes in the environment. Our objective is to develop cognitive multi-spectral sensors that directly generate trustworthy insights from wideband multi-modal analog signals using closed-loop feedback control of the sensor hardware and feature extraction algorithms that enable energy-efficient sensing-to-action. In an autonomous system, our cognitive sensor will make efficient use of the available sensor power by only sampling the most useful part of a scene and generating only the most useful features. We will pursue a full-stack approach connecting circuits; signal processing, machine learning, and control algorithms; heterogeneous integration; system software; and applications to develop and demonstrate the COGNISENSE concepts. The center will be organized in five interacting themes: (1) Multi-spectral Array, (2) Analog-to-Insight, (3) Closed-loop Attention Control, (4) Heterogeneous System Integration, and (5) System Software and Integration.