MANETs网络基于Q学习和网络编码的传输技术研究

发布时间：2018-05-03 06:27

本文选题：移动自组织网络 + 传输策略　；参考：《电子科技大学》2014年硕士论文

【摘要】：在移动自组织网络中,根据某一路由准则(比如跳数),源节点或中继节点通往目的节点的下一跳节点往往不只有一个可以选择,缺省采取的方法是随机选择下一跳节点,但这种选择方式会引起一些节点的拥塞增大端到端时延并且浪费网络资源。本文提出了基于Q学习算法的MANETs网络分组传输策略来合理选择通往目的节点下一跳节点,Q学习算法通过执行动作并且获得执行动作以后获得的奖赏值来更新相应动作的Q值,并基于Q值和一定的动作选择策略来决定下一跳节点。本文在OPNET环境下进行算法实现以及性能仿真,同随机选择下一跳节点的传输策略做端到端时延性能上的对比,通过对比可以发现基于Q学习算法的MANETs网络并发分组传输策略的端到端时延要明显低于随机选择下一跳节点传输策略的端到端时延。本文首先采用了基于当前Q值的贪心动作选择策略,在进行性能分析时发现其存在的问题:在单次的动作选择中只参考当前的Q值而并不考虑以往的经验,可能做出错误决定。因此本文又提出了基于Q值平均值的贪心动作选择策略,这样在执行动作选择时就可以从全局角度对比可选动作的优劣。同时基于Q平均值的贪心动作选择策略在时延性能上要比基于当前Q值的贪心动作选择策略的Q学习算法更加稳定。另外一方面,在存在干扰的场景下如何高效、正确地进行数据传输也是一个需要解决的问题。本文引入了近年来的研究热点—网络编码的思想提出了基于随机线性网络编码的端到端传输策略,并在OPNET环境下协同MATLAB加以实现,同停等ARQ协议做了仿真性能的对比,通过对比发现基于随机线性网络编码的端到端传输策略在传输有效性上要优于停等ARQ协议。最后本文对编码块的大小对传输性能的影响做了研究,发现编码块较大数据传输的实时性差,但是有效比特率高,编码块较小数据传输的实时性好,但是有效比特率低。
[Abstract]:In a mobile ad hoc network, the next hop node from the source node or relay node to the destination node often has more than one choice according to the criteria (such as hops), and the default method is to select the next hop node randomly. However, this method can cause congestion of some nodes to increase end-to-end delay and waste network resources. In this paper, a MANETs network packet transmission strategy based on Q learning algorithm is proposed to reasonably select the Q value of the next hop node to the destination node by performing the action and obtaining the reward value obtained after the action is performed to update the Q value of the corresponding action. And based on Q value and certain action selection strategy to determine the next hop node. In this paper, the algorithm implementation and performance simulation are carried out in OPNET environment, which is compared with the end-to-end delay performance of random next hop node transmission strategy. It is found that the end-to-end delay of the concurrent packet transmission strategy in MANETs networks based on Q learning algorithm is significantly lower than that of the next hop node transmission strategy selected at random. In this paper, the greedy action selection strategy based on the current Q value is first adopted, and the problem is found in the performance analysis: in a single action selection, only the current Q value is referred to and the previous experience is not taken into account, and the wrong decision may be made. Therefore, a greedy action selection strategy based on the average Q value is proposed in this paper, so as to compare the advantages and disadvantages of the optional action from a global perspective when performing the action selection. At the same time, the greedy action selection strategy based on Q average is more stable than the Q learning algorithm based on the current Q value based greedy action selection strategy. On the other hand, how to transmit data efficiently and correctly in the presence of interference is also a problem to be solved. In this paper, an end-to-end transmission strategy based on stochastic linear network coding is proposed by introducing the idea of network coding, which is a hot research topic in recent years. The scheme is implemented by cooperative MATLAB in OPNET environment. The simulation performance of ARQ protocol is compared with that of ARQ protocol such as stop. It is found that the end-to-end transmission strategy based on stochastic linear network coding is superior to the stopping ARQ protocol in terms of transmission efficiency. Finally, the effect of the size of the coding block on the transmission performance is studied. It is found that the real time performance of the larger data transmission block is poor, but the effective bit rate is high, and the real time performance of the coding block smaller data transmission is good, but the effective bit rate is low.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN929.5

【参考文献】