【作者】 刘微；

【导师】 陈贺新；

【作者基本信息】 吉林大学 ， 通信与信息系统， 2011， 博士

【摘要】 在“两化融合”和“感知中国”的国家战略背景下,物联网发展受到了政府、产业、资本等各层面的高度关注,射频识别(Radio Frequency Identification, RFID)技术作为物联网的主要驱动技术,已被列为本世纪十大重要技术之一。目前,RFID系统在物流、交通和零售等领域形成了小规模的市场,但其自动化、智能化、协同化程度仍然较低,其应用基础技术研究还存在着大量尚未解决的关键问题。而RFID系统优化技术作为保障RFID系统稳定、可靠和安全运行的基础,已成为现阶段RFID技术研究与应用的重要课题。本文根据RFID系统优化一般为非线性、多目标、大规模的复杂优化问题,利用智能算法求解这类问题时在计算精度、收敛性、初值敏感以及解的鲁棒性和自适应性等方面较传统数学优化算法更具优势的特点,在综述生物启发式计算研究的基础上,提出基于生物行为的RFID系统优化模型与算法。旨在通过深入研究通用、可扩展的RFID系统优化模型,设计一整套高效、可靠的基于生物行为的智能优化算法,重点解决RFID实际大规模应用中读写器调度、网络负载均衡、标签覆盖以及多读写器数据融合等相关优化问题,以提高RFID系统的运行效率和服务质量。论文的主要研究内容包括以下4个方面：1.研究了基于多种群共生粒子群优化算法(Symbiotic Multi-Species Particle Swarm Optimizer, SMPSO)的RFID读写器防冲突问题。在分析RFID读写器冲突及其建模问题的基础上,研究了考虑最小化读写器冲突和总处理时间的RFID读写器防冲突优化模型；在标准粒子群算法的基础上,基于自然界中的生物共生理论提出了SMPSO算法。SMPSO算法通过定义单物种内协作与物种间的信息交流机制,建立生态系统中的互利共生策略,具有更好的多样性保持能力及后期搜索性能。将基于SMPSO的RFID网络防冲突算法应用于四个不同规模的RFID读写器网络进行仿真实验,通过与标准粒子群(Particle Swarm Optimization, PSO)算法比较迭代过程中的进化曲线,表明SMPSO算法的收敛速度和解均优于标准PSO算法。仿真实验表明SMPSO算法能够有效求解密集读写器环境下的读写器冲突问题,并优化整个读写器网络的工作效率。2.提出了多群体协同人工蜂群聚类方法,并将其应用于RFID在制品跟踪系统数据的分析和处理。将多群体协同进化模型引入到人工蜂群优化(Artificial Bee Colony,ABC)算法中,提出了多群体协同人工蜂群优化算法(Multi-Species Cooperative Artificial Bee Colony Optimization, MCABC),将MCABC算法与ABC、PSO以及协同粒子群算法(Cooperative Particle Swarm Optimization, CPSO)在3个标准聚类问题数据集上进行了实验仿真,仿真结果表明MCABC算法的收敛速度、求解的精度和结果鲁棒性均优于其它算法；将基于MCABC算法的聚类分析应用于RFID在制品跟踪系统的数据处理模块,提出了基于MCABC聚类的RFID在制品跟踪数据处理模型,仿真实验结果表明该模型可以有效地识别和剔除错误观测值,提高RFID读写器网络目标定位跟踪的精度。3.研究了基于层次菌群觅食优化(Hierarchical Bacterial Foraging Optimization, HBFO)算法的RFID网络规划。提出了通用、可扩展的RFID系统优化模型,设计了基于层次群体智能优化模型的HBFO算法,并将其应用于RFID网络规划以处理大规模、多目标的复杂多峰问题。通过对HBFO、遗传算法(Genetic algorithm, GA)、PSO算法求解RFID网络规划的仿真实验结果对比,表明HBFO算法在标签覆盖、读写器干扰、网络负载均衡分目标以及网络经济效率总体目标函数的优化结果均优于其它两种优化算法。4.设计了自适应细菌觅食优化(Self-Adaptive Bacterial Foraging Optimization, SABFO)算法,并将其应用于动态RFID网络规划。将自适应搜索策略和群体感应机制引入基本细菌觅食优化(Bacterial Foraging Optimization, BFO)模型,建立SABFO算法模型。对SABFO、PSO、BFO及实数编码的GA算法在一组标准测试函数中进行仿真实验,仿真结果表明SABFO算法解的收敛速度、解的精度较其它算法有不同程度的改善。根据RFID系统本身的动态性和不确定性,在静态RFID网络规划模型基础上建立了RFID网络动态优化模型。通过对BFO算法与SABFO算法求解动态RFID网络规划模型进行仿真实验,结果对比表明综合考虑标签覆盖和读写器干扰的总体目标函数优化的SABFO算法优于BFO算法。仿真结果表明,SABFO算法能够持续地追踪到变化的峰值,适用于求解动态工程优化问题。更多还原

【Abstract】 Under the background of national strategy-Digital Convergence and Sensing China, Internet of things has received great attention from the government, industry, stock, and ect. RFID technology, as the main driving technology of the Internet of things, is considered one of the top ten important technologies in this century. At present, the applications of RFID system form a small scale market in the logistics, transportation, retail, and etc. However, the automation, intelligence, and coordination of the RFID system are still in low level. There are a lot of key problems of foundamental application technical researches to resolve. Among them, the RFID system optimization technique, as the basis to enssure the RFID system stable, reliable and safe operation, has become important issues in the RFID technology research and application.Due to the RFID system optimization generally being nonlinear, multi-objective, and large-scale complex problems, this thesis uses the characteristics that the intelligent optimization algorithms have more advantages than the traditional mathematical optimization algotithms in accuracy, convergence, initial value sensitivity, robustness and adaptability of solutions, and ect., for solving this kind of problems. On basis of the review of biological heuristic calculation researches, RFID system optimization models and intelligent altorithms based on biological behaviors are proposed. Through in-depth studing general and extensible RFID system optimization model, and designing a set of efficient and reliable intelligent optimization algorithms based on biological behaviors, the research fruits can mainly solve the optimization problems related to RFID reader scheduling, network loading balance, labels coverage, and multi-readers’ data fusion, and ect., which improve the operation efficiency and service quality of the RFID system.The main research contents include following four aspects:1. The RFID reader anti-collision problems are studied by Symbiotic Multi-Species Particle Swarm Optimization(SMPSO). Through analying the RFID reader collision and conflicting models, a RFID reader anti-collision optimization model is studied, which considers minimizing reader conflicting and total processing time. Based on the standard particle swarm algorithm (PSO), the SMPSO algorithm, on account of biological symbiosis theory of the nature, is presented. By defining the information communication mechanism of cooperation in single species and collaboration between species, symbiotic strategy in ecological system is established, which has better diversity keeping ability and later searching performance. Usillizing simulation experiment that applies the SMPSO and PSO to four different sizes of RFID reader networks, the simulation result is showed that SMPSO’s convergence speed and solutions are superior to that of PSO through comparing evolutionary curve during the iterative process. The Simulation experiment result shows that the SMPSO algorithm can effectively solve reader collision problem in intensive reader environment, and optimize the efficiency of the whole reader network.2. A clustering method is put forward based on Multi-Species Cooperative Artificial Bee Colony(MCABC) optimizaiton algorithm, that is applied in data analysis and processing of the RFID products in process tracking system. Taking advantage of multi-species co-evolutionary model, MCABC algorithm is proposed. A simulation experiment is tested on three standard clustering data set by MCABC, Artificial Bee Colony(ABC), PSO, and Cooperative Particle Swarm Optimization(CPSO). The simulation result shows that the convergence speed, solution accuracy and robustness of MCABC are superior to the other algorithms. The MCABC clustering algorithm is applied to the data processing module of the RFID products in process tracking system, then the data processing optimization model for the tracking system based on MCABC is propose. The simulation result shows that the model can effectively identify and eliminate errors, that may imporve the locating and tracking target accuracy of the RFID reader networks.3. The RFID network planning based on Hierarchical Bacterial Foraging Optimization (HBFO) is researched. A general and extensible RFID system optimization model is presented, then the HBFO algorithm based on heirarchy swarm intelligence optimization is designed, which is applied in the RFID network planning to solve large scale, multi-objective complex multi-peak problems. By the simulation experiment that HBFO, Genetic algorithm(GA), and PSO is used to solve the RFID networds planning, the simulation result reveals that the HBFO algorithm is superior to the other two algorithm in single objective function of labels coverage, reader interference, network loading balance and overall objective function of network economic efficiency.4. The Self-Adaptive Bacterial Foraging Optimization (SABFO) algorithm and RFID network dynamic optimization model are designed. Utilizing adaptive searching strategy and quorum-sensing mechanism,introduced to basic Bacteria foraging optimization (BFO) model,SABFO algorithm is built. By simulation experiment that is tested by applying the SABFO, PSO, BFO,and GA based real-coding algorithms in a group of standard test functions, the simulation results show that the convergence speed and solution accuracy of the SABFO are improved by different degrees.According to the dynamic and uncertainty of the RFID system itself, the RFID network dynamic optimization model is established based on static RFID network planning model, A simulation experiment is set through更多还原