Abstract

The thesis is motivated by recent advancements and developments in large, distributed, autonomous, and self-aware Cyber-Physical Systems (CPSs), which are emerging engineering systems with integrated processing, control, and communication capabilities. Efficient usage of available resources (communication,computation, bandwidth, and energy) is a pre-requisite for productive operation of CPSs, where security, privacy, and/or power considerations limit the number of information transfers between neighbouring sensors. In this regard, the focus of the thesis is on information acquisition, state estimation, and learning in the context of CPSs by adopting an Event-based Estimation (EBE) strategy, where information transfer is performed only in the occurrence of specific events identified via the localized triggering mechanisms. In particular, the thesis aims to address the following identified drawbacks of the existing EBE methodologies: (i) At one hand, while EBE using Gaussian-based approximations of the event-triggered posterior has been fairly investigated, application of non-linear, non-Gaussian filtering using particle filters is still in its infancy, and; (ii) On the other hand, the common assumption in the existing EBE strategies is having a binary (idle and event) decision process where during idle epochs, the sensor holds on to its local measurements while during the event epochs measurement communication happens. Although the binary event-based transfer of measurements potentially reduces the communication overhead, still communicating raw measurements during all the event instances could be very costly. To address the aforementioned shortcomings of existing EBE methodologies, first, an intuitively pleasing event-based particle filtering (EBPF) framework is proposed for centralized, hierarchical, and distributed (iii)state estimation architectures. Furthermore, a novel ternary event-triggering framework, referred to as the TEB-PF, is proposed by introducing the ternary event-triggering (TET) mechanism coupled with a non-Gaussian fusion strategy that jointly incorporates hybrid measurements within the particle filtering framework. Instead of using binary decision criteria, the proposed TET mechanism uses three local decision cases resulting in set-valued, quantized, and point-valued measurements. Due to a joint utilization of quantized and set-valued measurements in addition to the point-valued ones, the proposed TEB-PF simultaneously reduces the communication overhead, in comparison to its binary triggering counterparts, while also improves the estimation accuracy especially in low communication rates.