有环网络CNC译码算法优化设计

Design of Decoding Algorithm Optimization for Cyclic Network

摘要: 卷积网络编码(CNC)译码算法汇点容量区域及译码算法已知的假设条件并不能同时成立。为此在分析有环网络CNC译码机制的基础上，指出解决CNC译码问题的关键在于初始化算法的设计和容量区域确值问题。通过简化系统状态方程，建立发送和接收序列关系表达式，区分节点是否具有重置/清零操作，差异化提出与之对应的译码算法，并推导传输矩阵满秩条件时容量区域定量表达式，并对比分析了时延、吞吐量、丢包率、控制开销等性能指标。结果表明，新算法吞吐量提高了3.2 Mbps，时延、丢包率、控制开销分别降低了37%、25%、0.9 M，验证了新算法的可行性和有效性，为进一步提高网络安全性提供了一定理论基础。

Abstract: Decoding procedure of convolutional network coding (CNC) relies on multiple dimension prior information of the given network. Moreover, two assumptions, that are, known capacity region and decoding algorithm, couldn’t be established simultaneously. Therefore, after analyzing the current decoding mechanism of CNC in cyclic network, algorithm optimization should be focused on two key factors, called as initialization algorithm design and capacity region confirmation. Thus a simplified state space representation of the system was founded, which established an input-output relationship. And two differential decoding algorithms were proposed after distinguishing whether nodes have resetting/clearing operation or not. Then a quantified expression of capacity region was derived based on full rank matrix of transfer function. At last, network performance was analyzed by delay, throughput, packet loss rate and control overhead. Results show that, for the novel algorithm, throughput raises about 3.2 Mbps, delay reduces by 37%, packet loss rate decreases by 25%, and control overhead diminishes about 0.9 M, all of which prove that the proposed algorithm is feasible and efficient, providing a theoretical basis for better improving the security level of the network.