NCSs have become quite popular recently due to their growing applications in industrial automation, smart transportation, remote robotic control, etc.. A typical NCS consists of a MIMO dynamic plant, multiple wireless sensors and a controller. These are connected over a wireless communication network and form a closed-loop control. Conventional wireless networks are mostly designed to support delivery of content and information. However, NCS applications pose very different challenges to the design of wireless networks, which distinguish them from traditional wireless system designs. In this thesis, we explore novel ways of designing the wireless networks that are tailored for NCS applications. Regarding the PHY layer design, we propose a closed-form fully dynamic MIMO precodi...[ Read more ]

NCSs have become quite popular recently due to their growing applications in industrial automation, smart transportation, remote robotic control, etc.. A typical NCS consists of a MIMO dynamic plant, multiple wireless sensors and a controller. These are connected over a wireless communication network and form a closed-loop control. Conventional wireless networks are mostly designed to support delivery of content and information. However, NCS applications pose very different challenges to the design of wireless networks, which distinguish them from traditional wireless system designs. In this thesis, we explore novel ways of designing the wireless networks that are tailored for NCS applications. Regarding the PHY layer design, we propose a closed-form fully dynamic MIMO precoding and power allocation solution which adapts to the complete system state (MIMO channel state for transmission opportunity, sensor energy state for energy availability and plant state estimation error for the urgency of the state transmission). These agile adaptivities are very important for superb performance in the NCS applications. For the MAC layer design, we propose a novel modulation-free MAC protocol for NCS applications. The proposed scheme exploits interference and collisions among active sensors in the wireless channels to enhance the performance of remote state estimation rather than avoiding collision. As such, all sensors in one control domain can share a common spectrum and achieve zero MAC latency regardless of the number of sensors involved. This is very different from the traditional paradigm of communication network design for content delivery.

Besides the innovations on the NCS system design, we also propose a systematic NCS stability analysis framework. The proposed framework provides closed-form characterizations of the stability condition for general measurement matrices, and explicitly quantifies the penalty of rank-deficient measurement matrices on NCS stability. At last, we point out that the stability of average state estimation MSE, which is adopted in most of the existing literature is not sufficient to characterize the estimation performance because it is only a crude characterization on the first order moment of the random MSE process and contains no information on the MSE sample trajectory realizations. We propose a new performance metric of MSE tail probability, which guarantees fine-grained characterizations of the MSE sample path. We also developed various sufficient conditions for the well-behavedness of the tail distribution of the MSE.