WSAN中时延约束的协作数据汇聚能效优化研究

发布时间：2018-04-28 00:33

本文选题：无线传感器-执行器网络 + 能效优化　；参考：《中南大学》2014年博士论文

【摘要】：摘要：能效优化是无线传感器-执行器网络(Wireless Sensor and Actuator Network, WSAN)的核心问题之一。近年来,以动态拓扑控制和节点分布式协作为基本策略的协作数据汇聚是提升WSAN能量效率的重要研究方向。本文以重载桥梁和路基长期安全状态监测的WSAN为研究背景,采用增强学习、协作通信和博弈论等理论和方法,研究执行器路径规划、多执行器负载平衡和节点干扰抑制等问题,提出WSAN中时延约束下的能效优化机制。主要工作包括以下几个方面：论文从应用于监测重载桥梁和路基的WSAN实例出发,分析了时延约束下多执行器WSAN能效优化的重点和难点,提出了基于移动数据汇聚和协作通信干扰抑制的WSAN系统架构,首先构建具有能量感知的动态拓扑,以此为基础选择轮询点,再规划执行器采集信息的路径；然后针对多执行器情况,采用在线模糊Q学习方法对网络进行自适应分区,保证多个执行器间的通信负载和能耗均衡；针对执行器采集信息过程中相邻节点间存在的数据干扰,提出基于协作通信和博弈论的干扰抑制方法。为实现低数据汇聚时延、高能量效率、高可靠性的WSAN提供一套有效的解决方案。提高网络能效,优化网络生存时间是WSAN的核心目标。论文针对多跳路由数据汇聚的负载均衡问题,选择网络数据汇聚轮询点。首先根据传感器节点的剩余能量构造网络动态生成树；在全局拓扑信息可得的场景下提出基于入度优先的轮询点选择算法,在只能获得局部拓扑信息的场景下提出基于节点通信负载的轮询点选择算法。选择的轮询点作为局部网络数据的汇聚点以分散化、区域化的方式限制网络的路由跳数,均衡网络能耗。针对WSAN的实时性要求,本文提出时延约束下的移动执行器的赛道寻优路径规划算法。以轮询点为执行器路径规划停留的参考位置,将执行器的路径规划问题转换为旅行商问题；考虑轮询点通信半径,采用基于赛道寻优的最近邻点启发式算法,求解执行器遍历所有轮询点的最优移动轨迹；同时将网络数据汇聚的时延约束转换成执行器的移动距离约束,作为算法的迭代收敛条件,保证网络数据汇聚实时性,并通过建立和求解空间域最优规划模型提供算法性能评估。对多执行器场景,论文提出一种在线模糊Q学习的能耗均衡自适应分区算法。将每个执行器作为具有学习和决策能力的智能体,实时获取网络中传感器节点的能量和分布状态信息；再使用模糊推理机制离散化位置和能量分布状态,构建Q学习的状态空间和Q函数；根据当前的网络能量分布状态选择具有最大回报值的分区中心位置,并产生对应的Voronoi划分,实现能耗均衡的分区自适应调整。 WSAN中突发数据导致的相邻节点间的数据发送干扰是影响数据传输效率的重要问题。本文提出了一种基于协作数据汇聚的干扰抑制机制。首先将协作通信技术引入WSAN,通过协作中继转发和协作干扰转发的方式提升节点的抗干扰能力；然后考虑到节点自私特性,引入频谱共享机制激励中继节点参与协作干扰抑制；采用斯坦伯格博弈方法对协作干扰抑制过程进行分析；再通过求解博弈均衡来选择最优的策略进行协作信息传输,得到一种干扰情况下的协作数据传输方法。该方法可抑制节点间的干扰影响,确保数据公平、合理的协作传输。仿真实验验证了算法的有效性。
[Abstract]:Abstract : Energy efficiency optimization is one of the core problems of Wireless Sensor and Network ( WSAN ) . In recent years , collaborative data convergence based on dynamic topology control and node distributed collaboration is an important research direction to improve the energy efficiency of WSAN .

Based on the example of WSAN applied to monitoring heavy - load bridge and subgrade , the emphasis and difficulty of energy efficiency optimization of multi - actuator WSAN under time delay constraint is analyzed . Based on the mobile data convergence and cooperative communication interference suppression WSAN system architecture , a dynamic topology with energy perception is first constructed , which can be used as the base to select the polling point and then plan the path of the actuator acquisition information .
then aiming at the multi - actuator situation , an online fuzzy Q learning method is adopted to adaptively partition the network , so that communication load and energy consumption balance among the plurality of actuators are ensured ;
In order to realize low data convergence time delay , high energy efficiency and high reliability , a set of effective solutions for WSAN with low data convergence time delay , high energy efficiency and high reliability is proposed .

In order to improve the energy efficiency of the network , optimize the network survival time is the core goal of WSAN . In this paper , the network data convergence polling point is selected according to the load balance problem of multi - hop routing data convergence . Firstly , the tree is generated according to the residual energy of the sensor node .
A polling point selection algorithm based on input priority is proposed in the context of global topological information , and a polling point selection algorithm based on the node communication load can be presented in a scenario where only local topological information can be obtained .

Aiming at the real - time requirements of WSAN , this paper proposes a path planning algorithm for the track optimization of a mobile actuator under the delay constraint . The path planning problem of the actuator is converted into a traveling salesman problem by taking the polling point as the reference position of the actuator path planning and staying .
considering the polling point communication radius , a nearest neighbor heuristic algorithm based on the track optimization is adopted to solve the optimal moving track of the executor through all polling points ;
At the same time , the delay constraint of the convergence of the network data is converted into the moving distance constraint of the actuator . As the iterative convergence condition of the algorithm , the real - time performance of the network data is guaranteed , and the algorithm performance evaluation is provided by establishing and solving the spatial domain optimal planning model .

In this paper , an adaptive partitioning algorithm for energy consumption in online fuzzy Q learning is presented in this paper . Each actuator is used as an intelligent body with learning and decision - making ability , and the energy and distribution state information of sensor nodes in the network are obtained in real time .
then the state space and the Q function of the Q learning are constructed by using the discrete positions and the energy distribution states of the fuzzy inference mechanism ;
according to the current network energy distribution state , and generating the corresponding voracious division so as to realize the partition self - adaptation adjustment of energy consumption equalization .

The data transmission interference between adjacent nodes caused by burst data in WSAN is an important problem which affects the efficiency of data transmission . In this paper , a kind of interference suppression mechanism based on cooperative data convergence is put forward . Firstly , the cooperative communication technology is introduced into WSAN , and the anti - interference ability of the node is improved by cooperative relay forwarding and cooperative interference forwarding .
then taking into account the self - private characteristic of the node , introducing a spectrum sharing mechanism to excite the relay node to participate in the cooperative interference suppression ;
The cooperative interference suppression process is analyzed by Steinberg game method .
The method can suppress the interference effect between nodes , ensure the fairness of data and reasonable cooperative transmission , and the simulation experiment verifies the effectiveness of the algorithm .

【学位授予单位】：中南大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN929.5;TP212.9

【参考文献】