Title：Networked Large-Scale Sensing over Complex Systems
Speaker：Prof. Shuguang (Robert) Cui（Fellow，IEEE）
Host：Prof. Ping Zhang
Time：June 5th,2015 (Friday) 9:00-11:00 AM
Location：New Research Building Room 501
Abstract：
Data intelligence is the core building block for any modern and future cyber-physical systems, and it involves three major aspects: data processing, data storage, and data communication. Interesting and challenging research problems could be formulated over the interactions among the above three aspects in the context of cyber-physical systems. In this talk, we focus on one such interaction between data processing and data communication, to solve a specific problem on networked large-scale sensing, where data processing has to be performed in a distributed fashion over a communication network. In particular, we seek good estimates of the randomly-varying state process in a dynamic cyber-physical system, at multiple distributed sensing nodes, each of them only having a partial observation of the overall state. We allow nodes to talk with neighbors defined over a communication graph, where we introduce a communication rate constraint on the average number of message exchanges allowed across the network per unit time. In a distributed Kalman filtering framework, we establish the consensus result to show that the respective error variance at each distributed node converges weakly in distribution. In addition, with large deviation analysis, we could show that such a distribution collapses to a Dirac measure (i.e., the error performance achieved by the ideal centralized Kalman filter) exponentially fast as we increase the network communication rate. To further satisfy more practical communication requirements, we then extend the result to the case with quantized message exchanges, with similar convergence results established. Towards the end of the talk, I will briefly mention another result on the interaction between data storage and data communication.
SKL-NST
June 1st,2015
