基于Q学习的多目标耦合协同任务分配算法

摘要

针对一般WSN多目标协同跟踪研究中优化目标单一的缺点,提出了一种基于Q学习的多目标耦合协同任务分配算法.该算法提出了合簇的方法以解决多目标耦合情形的网络任务分配竞争冲突问题,首先,采用Q学习方法选取多目标相遇阶段的最优合簇时机,同时拟定合适的Q学习函数和回报函数,得出目标并行阶段最优簇首切换方案;然后,在保证剩余能量的前提下,利用设计的信息效用函数和动态最小簇成员数目给出不同阶段最优簇首及簇成员的选择;最后,根据目标特征标签分离目标信息.仿真结果表明,算法可以对多目标跟踪的耦合情形进行优化,能够满足跟踪精度的需求,具有降低系统能量消耗的优点,较好地延长了网络的生命周期.%Aiming at the shortcoming of the singular optimization objective in the general research of the Wireless Sensor Network (WSN) multi-target cooperative tracking, a new multi-target coupling cooperative task allocation method based on Q learning is proposed.A cluster merging method is used to solve the competition conflict problem in task allocation in the case of multi-target coupling.Firstly, Q learning method is used to select the optimal time for merging clusters at the stage of multi-target encountering.At the same time, the suitable Q learning function and return function are found out, and the most suitable switching scheme of the cluster head for the stage of target parallel moving is obtained.Then, under the premise of ensuring sufficient remaining energy, the schemes to select the optimal cluster heads and cluster members of different stages are given by using the information utility function and the number of the smallest cluster members.Finally, the target information is separated according to the target tag.The simulation results show that the algorithm can optimize the multi-target coupling, which can satisfy the demand for tracking precision, and has the advantage of reducing the system energy consumption and prolonging the life cycle of the network.