未知拓扑无线自组网络多节点干扰决策算法

摘要

An algorithm focus on multi-nodes jamming decision is proposed to fulfill the need of interdicting information transmission in wireless Ad-hoc networks.Firstly,a model of Poisson point process (PPP) network is constructed in terms of the structure of wireless Ad-hoc networks,and then it is used to simulate the process of data transmission.Secondly,random interference to multiple nodes are performed,and the number of stopped network flows is counted by monitoring ACK information or reconnoitering nodes activities.A correlation matrix is constructed from jamming effects.Finally,the correlation matrix is continuously updated in jamming process by taking the advantage of interaction in time of reinforcement learning and is used for the selection of subsequent nodes.The proposed algorithm does not need to have a priori knowledge of information such as network topology or importance of nodes,and interaction is only needed in learning nodes correlation matrix which would be helpful when choose nodes to jam.Simulation results by jamming wireless Ad-hoc in different circumstances and a comparison with the joint slotted exploit explore learning show that the proposed interdiction algorithm increases by 27.1％ in accumulate stopped flows,and its robustness is superior to existing algorithms.%为满足战场环境下无线自组网络通信拒止的干扰需求,提出了一种未知拓扑无线自组网络多节点干扰决策算法(CUCB).首先,根据战场无线自组网络结构特点构建泊松点过程(PPP)网络模型,并利用其模拟网络中数据流传输过程;其次,随机对PPP网络中多个节点进行干扰,通过监听确认帧信息或侦察节点活跃度判断阻断网络流数,根据干扰结果构造节点相关性矩阵;最后,利用强化学习与环境实时交互的特点,在干扰过程中不断更新节点相关性矩阵并将其用于后续节点选择.所提算法无需获悉目标网络拓扑结构、节点重要性等先验信息,仅以阻断网络流数目或节点活跃性作为奖赏标准,适用网络类型更为广泛.仿真结果表明,对不同参数下的无线自组网络进行干扰,所提算法在鲁棒性方面优于现有算法,在累积阻断网络流数量方面比联合利用探索算法提高了27.1％.