On Performance Analysis And Control Of Networked Systems

Networked systems are commonly seen in real world. All the things in both macrocosmos and microcosmos can be described and analyzed from the point view of networked systems. The discussions in this dissertation refer to two representative networked systems:networked control systems and wireless sensor networks. Reference input, plant output and control input signals are transmitted through networks between controllers, actors and sensors, which makes networked control systems more flexible with more powerful function. Networked control systems make it possible to share the resources, to reduce the complexity and producing costs, and thus, attract many attentions from researchers in recent years. Networks bring convenience to control systems, at the same time, it bring new challenges. Generally speaking, signals that transmitted in networks will also suffer the constraints originated from communication, which affect stability and performance of networked control systems. Recent studies are focused on stability of systems with considering networks induced parameters, such as time delay based on state space method. However, there are few works on the study of networked control systems performance based on frequency optimal control method. This dissertation addresses issues in this spirit, it try to establish the relationship between communication channel characteristic parameters and control systems performance through the analysis of limitations on networked control systems performance. Also, this dissertation studies another kind of networks and systems, which is wireless sensor networks. The studies on wireless sensor networks relate to knowledge in the field of electronic circuit, computer science, communication and control theory. Nowadays, the study on wireless sensor networks can be seen in many fields. From the view of complex networks, this dissertation establishes models of wireless sensor networks and gives the analysis by the tools that are commonly seen in complex networks analysis, which gives a way to solve the conflict between networks performance and limited nodes energy, and thus improves lifetime of wireless sensor networks. The main contents of this dissertation are outlined as follows.Averaged tracking performance limitations in linear time invariant multi-input multi-output systems with random signal reference input has been studied. The random input signal considered is the Brownian motion sequence. Results show that tracking performance limitations in these kinds of systems depend on the structure of plant and statistic characteristics of reference input. Here the structure of plant refers to location and direction of nonminimum phase zeros and unstable poles. As special cases, this dissertation also studies tracking performance limitations when uniform random reference input and only single non-minimum phase zero with single unstable pole are considered. Finally, tracking performance limitations of two-degree-of-freedom controller tracking system with random reference input are given.This dissertation studies the tracking performance limitations of linear time invariant, multi-input multi-output, continuous-time systems in which the output feedback is subject to an additive white Gaussian noise corruption. The problem under consideration amounts to determining the minimal error in tracking a Brownian motion random process, which emulates a step reference signal in the deterministic setting. Both the unity feedback and two-degree-of-freedom controller structure are considered. In the former case this dissertation derives an explicit bound, and in the latter an exact expression of the minimal tracking error attainable under the noise effect. Both results demonstrate how the additive white Gaussian noise may degenerate the tracking performance, and how the noise effect may intertwine with unstable poles and nonminimum phase zeros which are intrinsic characteristics of the plant.Also, this dissertation considers the issue of tracking performance limitations for linear time invariant discrete-time systems when the output feedback signal is subject to uniform quantization. Based on a two-degree-of-freedom controller structure, this dissertation addresses the issue in both single input single output and multi-input multi-output systems. Results show that quantization invariably degenerates the achievable tracking performance. For a certain quantization interval, the extent of degeneration depends on the structure of the underlying system. For multi-input multi-output plant, it is found that a proper choice of the quantization method could reduce the negative effect of quantization, and thus improve the best tracking performance.Based on site percolation model, which is a special kind of random graphs, this dissertation establishes a model for wireless sensor networks. Relationship between work mode of node and connectivity of networks has been studied with the aid of phase transition phenomena in site percolation model, which reduces energy consumption of nodes and guarantees connectivity of networks. Furthermore, an algorithm has been proposed to mend the work mode of node. As a result, energy consumption becomes more averaged, and thus lifetime of networks is prolonged.A new kind of complex networks model named multi-radius geographical spatial network model has been proposed. This dissertation studies statistic characteristics of this model and then proposes an efficient mechanism of broadcasting radius adjustment that maps sensor networks to multi-radius geographical spatial networks. Analysis and simulation show that sensor networks working under our mechanism could consume energy uniformly to prolong lifetime of networks and have faster data delivering speed than those in traditional uniform radius sensor networks.Finally, a summary for all discussions is given in the dissertation. Future works that related to this work are also presented.