基于强化学习和蚁群算法的WSN节点故障诊断

为了克服现有的WSN节点故障诊断方法所具有的难以实现在线诊断和诊断精度仍然不够高的缺点,设计了一种基于Sarsa算法和改进蚁群算法的WSN节点在线故障诊断方法。首先,建立了监测区域的网络模型和WSN节点故障诊断模型,然后,采用主成分分析法对节点故障样本数据进行降维,从而提高诊断效率,将样本数据作为层次,将故障诊断类作为各层节点建立层次树,采用改进的Sarsa算法求取各层节点的Q值,并将其用于初始化蚁群算法中路径的信息素,最后,提出了一种改进的蚁群算法求取从第一层出发的蚁群到各层节点之间的路径,将各层中信息素最大的节点作为最终的故障诊断类别。在Matlab环境下进行仿真实验,结果证明文中方法能有效实现WSN节点故障诊断,且与其它方法相比,具有故障诊断精确度高且能在线故障的优点,是一种有效的节点故障诊断方法。

Fault Diagnosis Method for WSN Sensor Node Based on Reinforcement Learning and Ant Colony Algorithm

In order to conquer the problems such as difficultly diagnosing on-line and low diagnosis accuracy of the existing wireless sensor node fault diagnosis method, an on-line fault diagnosis method based on Sarsa algorithm and ant colony algorithm was proposed. Firstly, the network model and wireless sensor diagnosis model are built, then the principal component analysis is used to reduce the dimension of the fault diagnosis sample data to improve the diagnosis accuracy. The sample data is used as the layer and the diagnosis type is used as the node of the layer, then the layer tree is set up. The improved Sarsa algorithm is used to obtain the Q value of the node of every layer and then the Q value is used to initialize the pheromone of the route in ant colony algorithm. Finally, the improved ant colony is used to get the route from the ant colony from the first layer to every layer, and the node with the biggest pheromone is used as the final fault diagnosis type. In the simulation experiment in the Matlab, the result shows the method in this paper can implement the wireless sensor node diagnosing, and compared with the other methods, it has the higher diagnosis accuracy and on-line fault diagnosis. Therefore, it is an effective fault diagnosis method.