Sensor-Adaptive Signal Processing for Dynamic Decision Making

点击翻译按钮获取中文摘要

This project conducts fundamental research in adaptive decision making in signal processing systems. It involves the interplay of Bayesian signal processing, stochastic optimization, stochastic control and game theory. Such adaptive decision making problems have recently received significant attention in diverse areas including sensor networks, cognitive radars, biosensing, and automated decision systems. These problems transcend classical statistical signal processing (which deals with extracting signals from noisy measure- ments) to address the deeper issue of how sensors and signal processing algorithms can use feedback control to adapt to their environment thereby leading to a smart reconfigurable sensing systems. The research objectives of this project fall under the following three inter-related themes: (i) Sensor Resource Management: The objective is to devise adaptive resource allocation algorithms in sensor systems that are seeking to estimate the state of an underlying stochastic dynamical system. We are particularly interested in designing smart reconfigurable radars. (ii) Decentralized Decision Making: Motivated by energy, sensing and communication constraints, the aim is to devise autonomous strategies for individual sensors to activate themselves to detect events. Social learning and game theory will be used to synthesize and analyze decentralized decision making algorithms. (iii) Active Sensing using Molecular Biosensors: Our objective is to apply the fundamental theory, algorithms and analysis developed above to a remarkable biosensor that was built recently. The key unifying theme underpinning the above three objectives is the paradigm of sensor-adaptive signal processing: signal processing resources that adapt their behavior based on noisy measurements and previous decisions. Such adaptive decision making under uncertainty constitutes the fundamental theme of this project.