Fast Signal Processing and Control Algorithms for Complex Hierarchical Systems. Complex dynamical behaviour is inherent to many real-world systems including telecommunications networks, financial markets and biological systems. High performance signal processing and control algorithms for such large-scale, complex systems are computationally very expensive in general. An important class of large-scale Markovian models arising in many applications shows a remarkable hierarchical property, display ....Fast Signal Processing and Control Algorithms for Complex Hierarchical Systems. Complex dynamical behaviour is inherent to many real-world systems including telecommunications networks, financial markets and biological systems. High performance signal processing and control algorithms for such large-scale, complex systems are computationally very expensive in general. An important class of large-scale Markovian models arising in many applications shows a remarkable hierarchical property, displaying strong interactions within certain clusters of states and weak interactions among these clusters. By utilizing this property, the proposed project will design and analyze novel reduced-complexity signal processing and control algorithms with rigorous performance guarantees. In addition, this project will explore possibilities of making these algorithms hierarchical such that they are easy to implement through decentralization.Read moreRead less
Stochastic Sensor Scheduling in Statistical Signal Processing. In several statistical signal processing applications, due to computational or communication constraints, at each time instant one can use only a few out of several possible noisy (stochastic) sensors. The stochastic sensor scheduling problem deals with how to dynamically choose which group of sensors to pick at each time instant. This project involves research in sensor scheduling for widely used stochastic dynamical systems such as ....Stochastic Sensor Scheduling in Statistical Signal Processing. In several statistical signal processing applications, due to computational or communication constraints, at each time instant one can use only a few out of several possible noisy (stochastic) sensors. The stochastic sensor scheduling problem deals with how to dynamically choose which group of sensors to pick at each time instant. This project involves research in sensor scheduling for widely used stochastic dynamical systems such as Hidden Markov Models and Jump Markov Linear Systems. It focuses on the design and analysis of stochastic control algorithms such as dynamic programming and simulation based randomized methods. The research will lead to an integrated theory incorporating stochastic control, statistical signal processing and combinatorial optimization. We will also apply the resulting techniques to tracking maneuvering targets given noisy observations.Read moreRead less