WSN中传感器节点的弹性神经网络任务分配方法

为解决WSN多目标跟踪节点任务分配的竞争冲突问题,提出一种融合了模糊聚类的多弹性子模自组织神经网络节点任务分配方法.通过模糊聚类估计目标数量,建立节点任务分配跟踪精度和能量消耗的综合性能指标,采用非全连接的环形弹性结构自组织神经网络优化监测联盟,用最近邻法对神经元弹性子模进行初始化,根据胜者为王原则动态调整子模的感受域,以快速锁定最优监测联盟,实现多目标的精确跟踪.实验结果表明:文中方法能有效解决多目标跟踪节点任务分配的竞争冲突问题,以及竞争冲突时的系统能耗增加与实时性问题;在随机均匀部署节点拓扑和目标直线运动模式下,文中方法的能耗较最近邻法降低了48.2%~55.9%,较未改进弹性神经网络法降低了37.4%~42.5%,且运算速度提高了19.0%~27.4%.

Task Allocation Method of Sensor Nodes Based on MEMSOM Neural Network in WSN

Aiming at the competition conflict in node task allocation of WSN Multi-target tracking, the node dynamic alliance synergistic tracking method based on mixed fuzzy C means (FCM) and Multiple Elastic Modules (MEM) Self-Organizing Maps neural network is proposed. First, the target number is estimated by FCM. Second, combination performance index of tracking precision and energy consumption in multi-target tracking node task allocation is established. A self-organizing neural network with non-fully connected neuron ring elastic structure is constructed. Neurons in elastic submodule are initialized with the nearest neighbor method and neurons receptive field are dynamic adjusted with WTA (Winner Take All) principle in the attraction of task allocation combination property index, which makes node optimal alliance be quickly locked. Then the cluster heads are chose to run MTT program on energy principle and multi-target tracking can be achieved accurately. Experimental results show that the energy consumption based on improved MEM is reduced by 55.9% and 42.5% compared with the classical nearest neighbor method and the old MEM neural network method respectively, and calculation speed improves 27.2% compared with the old MEM neural network method. This method can effectively solve the problem of the competition conflict in MTT node task allocation and the increment of system energy consumption and real-time performance when dynamic coalitions compete and conflict for the resource of sensor nodes. The best compromise of system power, precision and real-time is achieved.