Stochastic sensor scheduling

点击翻译按钮获取中文摘要

Statistical signal processing deals with extracting signals from noisy measurements. In this project, motivated by energy, communication and computational constraints, we address the deeper issue of how to dynamically schedule and optimize signal processing resources to extract signals from noisy measurements. Such sensor scheduling/management problems involve the interplay of ideas in Bayesian signal processing, stochastic optimization, stochastic control and game theory. Our research involves four inter-related parts. In Part 1, we conduct fundamental research on the structural properties of sensor management problems formulated as constrained Markov decision processes, partially obeserved Markov decision processes and stochastic games. In Part 2, we exploit the structure in Part 1 to derive sophisticated simulation-based optimization algorithms for sensor management. In Part 3, we exploit the fundamental results in Part 1 and Part 2 to design and analyse efficient algorithms for networked sensors in surveillance and radar resource management. In Part 4, we use the fundamental results in Part 2 to devise efficient multiparticle Brownian dynamics and Molecular dynamics simulation of ion permeation in biological membrane ion channels. The key idea underpinning our research is the study of stochastic scheduling and games for dynamical systems. This unifying theme motivates our fundamental theory and applications.