Cochlea-inspired Wireless Compressive Sensing Architectures for Real-time Feedback Control Applications

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Feedback control systems are widely used to control the performance of dynamic systems found in mechanical (e.g., flight control of aircraft) and civil engineering (e.g., response mitigation of buildings during earthquakes). Wireless communication has been proposed as a communication medium for the sensors, actuators and controllers that make up the feedback control system. While wireless communication may dramatically reduce the costs associated with the manufacture of control systems, they simultaneously introduce a new set of technical challenges including power limitations, lower communication speeds and reduced system robustness. This project studies how the human cochlea and the auditory nervous system acquires, communicates and processes acoustic information with the aim of emulating its operational principles in wireless feedback control systems. The power-efficient and real-time processing architectures found in biological neural circuits promise to resolve the long-standing performance bottlenecks associated with wireless telemetry in feedback control applications. The impact this research would have on the US economy is enormous given the prevalence of feedback control systems used daily by society. In addition, the project introduces innovative engineering coursework in bio-inspired engineering and engages middle-school students from underrepresented groups to motivate them to pursue STEM careers. The overarching goal of the research effort is to apply the methods of neural pulse train coding from the field of neurobiology to resolve the power consumption, speed, and robustness challenges associated with wireless communication in networked feedback control systems. The team will study the sensory neural principles of the mammalian cochlea and will use these principles as the functional basis of an energy-efficient wireless sensor that modulates sensor data into digital pulse train signals communicated directly to the radio frequency spectrum using a code division multiple access (CDMA) communication scheme. The project will also study common neural motifs to serve as functional blocks for the mapping of feedback control laws into complex neural circuits that process the spike train signals generated by the cochlea-inspired wireless sensor nodes. The result will be a biologically inspired wireless feedback control system that implements a compressive sensing strategy at the sensing front-end and processes spike train signals in real-time on the system back-end using neural circuits trained to implement desired feedback control laws. The intellectual significance of the work is that it breaks new ground by adopting the operational principles of biological neural circuits to formalize a new design paradigm for engineered feedback control systems.