Executive Summary : | Cyber-physical systems (CPS) have gained significant interest due to their reliance on remote monitoring and control through inference. However, the complex interaction between inference, control, sensing, communication, and computation makes it computationally challenging to jointly optimize network operations and inference/control. To address this, a thorough understanding of the impact of network performance on inference/control performance is needed. Current literature does not help in understanding the combined effect of all network operations on networked inference/control or identifying key network performance metrics. The proposed work aims to study performance optimization of networked inference/control through three steps: solving networked inference/control for a given network operation, identifying key network performance metrics for inference/control performance, and designing network operations for these metrics. The primary case studies will include change point detection, networked estimation/tracking, and networked control. The proposed work will develop algorithms for quickest change detection of a Markov chain, optimize centralized and distributed tracking performance over estimation schemes, sampling, scheduling, and routing over a multihop wireless network, and develop optimal control rules under asynchronous and out-of-sequence packet delivery over a network. The optimization of age-of-information jointly over sampling, scheduling, and routing will be given special emphasis. The developed theory and algorithms will provide insights for CPS system design. |