基于梯度和模糊神经网络的容迟网络路由算法

DTN Routing Algorithm Based on Gradient and Fuzzy Neural Network

高效的路由算法是保证容迟网络性能的关键技术.为提高适用于容迟网络的路由算法的性能,提出了一种基于梯度和模糊神经网络决策的容迟网络路由算法.该算法具有如下特点:改进了网络描述向量,采用节点自身信息及节点间链路状态信息来描述网络,实现对网络的全面描述;将有限历史信息的动态平均与精确预测相结合,自适应维护网络描述向量的各分量,进而为路由决策提供准确的量度;采用模糊径向基神经网络进行路由决策,实现路由决策过程的智能化;依据多跳传输成功概率引导分组沿梯度方向转发,提高分组转发效率.仿真结果表明,在同等网络条件下,该算法表现出比传染路由算法和下文感知路由算法更优异的网络性能.

Efficient routing algorithm is the key to guaranteeing the performance of delay-tolerant network（DTN）.In order to improve the efficiency of DTN routing algorithms,a routing algorithm based on gradient and fuzzy neural network decision is proposed.This algorithm improves the vector for network description by adopting the own information of a node and the link state information between two nodes,which enables a comprehensive description of the network.Moreover,it provides accurate measurement for routing decisions by combining the dynamic average of limited historical information with an accurate prediction and by adaptively maintaining the components of the vector.In addition,fuzzy RBF（Radical Basis function） neural network is employed in the algorithm for routing decision so as to result in an intelligent routing decision-making process,and,the successful multi-hop packet transmission probability is used to guide the forwarding along the gradient,thus improving the packet forwarding efficiency.Simulated results indicate that the proposed algorithm achieves better performance than epidemic and context-aware routing algorithms under the same network conditions.