基于群集智能的传感器管理方法研究

摘要

The conventional sensor management approach generally uses centralized data processing to deal with the system information. The centralized approach has several issues, such as making the center node overloading, the communication of system being congregated, and it probably will bring out the problem of " tragedy of the commons" because of the introduction of local heuristic information. Against these problems, we suggest to solve the sensor management problem by COIN ( collective intelligence) theory. The method of sensor management based COIN proposed in this paper uses efficiency function as sensor utility function. In the sensor management method based COIN, the private energy function is defined for every sensor, while global energy function is defined for whole system. The iterative process of COIN theory ensures the global utility will be best when all of the individuals maximize its private utility.The simulation show that the method proposed in this paper can not only reduce the load of the central node and the communication pressure of system, but also can improve the system performance as well as the other sensor management methods. The compare with the sensor management method based PSO shows that the effect of sensor management proposed in this paper is better than that of PSO method.%传统传感器管理方法一般采用集中式处理,这种处理方式可能会产生中央节点负荷重和系统通讯压力过大等问题,同时也会因为局部启发式信息的引入而导致了公共悲剧问题.针对以上问题,本文将群集智能理论应用于多传感器管理研究,在以效能函数作为传感器效用函数的基础上,提出了基于群集智能的传感器管理方法.该方法为每个传感器定义个体效用函数,为整个传感器系统定义全局效用函数.群集智能理论保证了在算法迭代过程中通过优化传感器的个体效用而达到优化全局效用的目的.仿真实验表明使用群集智能理论解决传感器管理问题不仅可减轻中央节点负荷、降低系统通讯压力,在提高系统性能的方面也不逊色于其他传感器管理方法.以虚拟传感器网络数据作为实验数据的实验结果表明,本文所提出的方法在传感器的管理性能上,效果优于基于粒子群的传感器管理方法.