Dr. Munagala will develop theoretical models for the problems of uncertainty and imprecise information in computing and communication systems such as wireless sensor networks and grid computers, and design algorithms and policies to ameliorate their impact on system performance, thereby leading to more effective larger-scale deployments of such systems.
He will also develop algorithms for computing dynamic information acquisition strategies in uncertain environments, that jointly optimize the cost of acquiring more information, together with the gain from exploiting the information to enhance system performance. This class of problems falls within the larger area of stochastic control theory.
It is well known that computing and storing the optimal control policies (or strategies) needs exponential time and space. However, in the scenarios that Dr. Munagala will investigate, the system can tolerate certain suboptimality in the quality of planning decisions, but the real requirement is that the policies be extremely lightweight to compute and execute. This motivates the formulation and design of computationally efficient approximately optimal policies for an entire gamut of fundamental control theoretic problems.
The PI will draw upon his expertise in approximation and online algorithms to develop new insights to address important theoretical questions in stochastic control theory, with emphasis on information acquisition and exploitation problems. In contrast with techniques developed in the AI and operations research communities, the solutions developed will have provable performance guarantees, in addition to being efficient to implement. The novel algorithmic principles emerging from this project will lay the foundations for developing a fresh theory of approximate stochastic control.
