基于混合编码遗传算法的证据网节点可靠性评估

摘要

证据网是一种基于D-S（Demspter-Shafer）理论层次化推广的推理模型，和D-S理论一样，当证据网中传感器节点不可靠时需要进行折扣（可靠性）处理。由于证据网是一种多层次的节点信息融合，折扣在不同融合层次传感器上，影响不同层次上的冲突，所以折扣的设置需全局考虑冲突情况。已有的可靠性评估方法是 D-S理论中的评估，这些方法并不能保证在融合中全局的冲突最小，针对这一问题提出一种以减小全局冲突为目标使用混合编码遗传算法进行可靠性评估的方法。在仿真实验中通过与已有的可靠性评估方法进行比较，证明了该方法更能减小全局冲突，获得更好的结果。%Evidential Network is a reasoning model based on extending the Demspter-Shafer(D-S)theory, when the sensors node are unreliable,discounts(reliability)should be set at the nodes as in the D-S theory.Be-cause the evidential network fuses the multiple echelons of information of nodes,the discounts of sensors are at the different fusion levels and take effect at the different fusion levels,the whole conflict should be taken into consideration when the discounts are set up.The existing ways of evaluating the reliability are used in the D-S theory,and they cannot ensure the minimum of the whole conflict when taking the fusing in the evidential net-work.In order to resolve the problem,the hybrid-code genetic algorithm is proposed to evaluate the reliability of evidential network nodes,with the aim of reducing the whole conflict.Compared with some other algorithms,the re-sults prove that the proposed algorithm takes some advantages in reducing the conflict and finding a better result.