基于PPO算法的链路不相交多路径路由优化研究

Research on Link-Disjoint Multipath Routing Optimization Based on the PPO Algorithm

摘要: 链路不相交多路径路由是当前网络优化的重要方向，传统路由算法在面对网络动态变化时存在适应性差、效率低等问题。本文提出了一种基于强化学习的链路不相交多路径路由算法，具体采用PPO (近端策略优化)算法。实验结果表明该算法具有良好的收敛性与稳定性，所选路径集合的奖励显著优于随机方法，在不同网络状态下均表现出较强的泛化能力与适应能力。

Abstract: Link-disjoint multipath routing is an important direction in current network optimization. Traditional routing algorithms often suffer from poor adaptability and low efficiency when facing dynamic changes in networks. This paper proposes a link-disjoint multipath routing algorithm based on reinforcement learning, specifically using the PPO (Proximal Policy Optimization) algorithm. Experimental results show that the proposed algorithm exhibits good convergence and stability. The reward of the selected path sets is significantly better than that of random methods, and the algorithm demonstrates strong generalization and adaptability under different network conditions.