Study On Consensus Of Time Synchronization Problem In Restricted Sensor Networks

Sensor networks(SNs),one of the distributed and networked systems,are formed by low-cost sensors that each generally communicates with others via inter-sensor wireless networking.SNs have advantages of being large-scale,self-organizing,high flexibility and cost-effective,and are widely employed in intelligent manufacturing,transportation,and electric power fields,etc.The individual clocks on the decentralized sensor nodes are necessary to keep networked synchronization in order to intertwine sensor schedul-ing and in-network data aggregating.Therefore,networked time synchronization is the research hotspot in SNs.The essence of time synchronization over SNs is an estimation problem based on the wireless network.Hence,a large number of distributed time synchronization algorithms based on parameter estimation schemes were proposed,and the consensus theory pro-vided different ideas for designing distributed estimation algorithms and control laws for clock state-space models of multiple sensor systems.Uncertainty and insecurity in the wireless networks with limited battery power of nodes are major constraints to time syn-chronization in SNs.However,the existing consensus-based time synchronization algo-rithms have some prominent problems,such as low energy efficiency,low safety factor,and less consideration paid to randomized synchronization in the sense of probability.To be specific,the following problems need to be resolved:Considering the latency time has the most significant impact on the synchronization accuracy,how to achieve highly energy-efficient consensus time synchronization in large-scale SNs;how to implement robust consensus time synchronization in dynamic SNs with rapidly changing connectiv-ity;when an SN is under multiple attacks,how to devise proactive defense mechanisms to ensure the effectiveness of the consensus-based time synchronization algorithm.In general,this thesis first studies the feasibility of two types of estimation schemes from the perspective of strict dispersed-type framework,and then focuses on the performance improvement of consensus-based approach in the areas of energy efficiency,robustness,and security.The main contributions of this thesis are given as follows.(1)For the feasibility of realizing global time synchronization based on the distribut-ed algorithm under the restrictions of wireless communications,two general mathemat-ical models of distributed time synchronization schemes are established.One is called a "relative clock synchronization" model of the "quasi-" distributed scheme.The other is distributed estimation and synchronization control for a class of discrete dynamical network systems.Start with the latter;the pedigree of first-order and second-order con-sensus timing protocols is analyzed,and their classifications are provided.Three indices are also defined to evaluate the key statistical characteristics of global synchronization behavior.It can be seen from the synchronization control protocol based on distribu-tion estimation and the construction process of synchronization control objectives that the stable control idea of the discrete dynamic network system is more feasible than the relative synchronization idea in terms of globally fully distributed framework.(2)The existing consensus-based time synchronization algorithms slowly converge to acceptable time synchronization precision over large-scale nodes.An optimal con-sensus time-synchronization algorithm based on topology deployment optimization is proposed to improve energy consumption.In order to minimize the message complexity and energy consumption,a subset of nodes along with a finite number of neighbors are selected in advance.Further,the selection of "subset" nodes should ensure connectivity for consensus propagation and the neighboring nodes should be assigned in such a way that the delay must be minimized and balanced for a faster convergence speed.Hence,a multi-objective optimization model is formulated by considering the balanced connect-ed dominating set and the magnitude of delay,which is proved to be NP-complete.In order to make the problem easy to solve,a multi-objective human learning algorithm by adding the dawdler strategy,is proposed to seek out Pareto solution,so as to obtain the topological connectivity indexes over the minimum delay,which will effectively reduce the energy consumption of the global network without compromising synchronization accuracy.(3)Gossip protocols are normally used for random geometric graph models to facili-tate the infrastructure-less communication of centerless networked systems.Considering the rapidly changing topology and the demand for peer-to-peer communication,an inno-vative consensus time synchronization algorithm based on randomized gossip interaction is proposed.The unreliable links are described by stochastic connections,reflecting the characteristic of changing connectivity gleaned from dynamic SNs.Based on the mutual drift estimation,each pair of activated nodes fully calibrates the clock rate and offset to achieve network-wide synchronization by drawing upon the gossip-based consensus ap-proach.The theoretical results on the probabilistic time synchronization performance are presented.Considering the problem of slower convergence speed,on the one hand,the converge-to-max criterion is introduced to achieve a much faster speed of convergence;on the other hand,the realization of the multi-gossip scheme based on edge dyeing is further proposed.Considering an asymmetric bounded delay model,the recursive least-squares estimator is designed to achieve the asymptotic consensus estimation of drift.This makes the revised gossip-based consensus algorithm become a more robust timing protocol.(4)The SNs are vulnerable to information manipulation attacks and Sybil attacks as they are deployed in hostile and unattended environments.Due to the fuzzy network boundary and the lack of physical protection of nodes,the security of gossip evolution faces challenges.To this end,a proactive defense strategy is applied to the gossip-based consensus time synchronization algorithm under the information manipulation attack and Sybil attack.First,the global logical clock error is proved to be non-convergence with probability 1 under Sybil attack.Second,based on time message correlation and clock frequency uniqueness,a trusted-node-based mechanism is designed to eliminate false information.Then it is introduced to the gossip-based consensus protocol,which can effectively eliminate multiple false data sources with low computing and storage load.Finally,the simulation results reveal that the convergence time of the proposed algorithm is not sensitive to attack frequency and can defend against both the message manipulation attacks and Sybil attacks.