Sensor systems are facing significant transmission bandwidth, energy, and data storage constraints caused by the vast amount of resources needed to maintain sensing accuracy in a dynamic environment. This situation is exacerbated when the sensor is highly mobile or when the sensor site is hostile, as it is very difficult to manually reconfigure a sensor to accommodate any dynamics in the environment. Current sensor designs are either non-adaptive or require large storage resources and, as a result, only a small fraction of the data be conveyed at a sufficient quality level to be useful for applications. If it is possible for the sensor system to intelligently adapt to the rapidly changing environment, then information from previously collected data can be used to instantaneously guide sensor configuration and the selection of data without the need to store and process the data offline. This project is a new neural-inspired approach to enable real-time adaptive sensing, such that the overall bandwidth usage, power efficiency, and sensor accuracy of the sensor system will be improved. Accurate real-time adaptive sensing can significantly reduce property damage due to natural and manmade disasters. It can also improve national security by identifying intentional terrorist actions, as well as providing an intelligent real-time large bandwidth adaptive sensing system for security applications. The educational impact of this work comes from its multi-disciplinary foundation, broadening students' views and encouraging them to think outside the box.
This project borrows the Crayfish Tail-flip Escape Response and the Spike Timing Dependent Plasticity from physiological circuitry, and aims to implement it with photonics. The focus of this project is to design and develop a photonic neural crayfish circuit, capable of real-time learning and adapting to rapid changes, and to incorporate the photonic crayfish circuit for real-time adaptive control in a sensor system. This initiative is an interdisciplinary integrated device and systems design study involving basic research from the standpoint of neuromorphic processing. The project aims to exploit physical processes in photonic devices, and to solve the problem of fundamental real-time adaptive control in a dynamic and complex modern sensor system. This fundamentally novel approach is a neural-inspired way to solve the problem by eliminating the computing bottleneck and providing high fidelities and large bandwidth processing ability that is needed for real-time adaptive sensing.
