| 基于强化学习的光传送网路由波长优化 |
| |
| 引用本文： | 孔英会,杨佳治,高会生,胡正伟.基于强化学习的光传送网路由波长优化[J].红外与激光工程,2022,51(11):20220084-1-20220084-9. |
| |
| 作者姓名： | 孔英会 杨佳治 高会生 胡正伟 |
| |
| 作者单位： | 1.华北电力大学 电子与通信工程系，河北 保定 071003 |
| |
| 基金项目： | 国家自然科学基金 (52177083)；河北省省级科技计划（SZX2020034）；国网山西省电力公司科学技术项目(SGSXXT00JFJS2100106) |
| |
| 摘 要： | 针对光传送网中动态业务的路由和波长问题，提出一种基于强化学习的深度路由波长分配算法DeepRWA。算法基于软件定义网络架构，通过强化学习灵活地调整控制光传送网，实现光网络路由波长分配策略优化。针对路由选择问题，结合链路上的波长使用情况，使用A3C算法选择合适的路由，使得阻塞率最小；针对波长分配问题，使用首次命中算法选择波长。考虑阻塞率、资源利用率、策略熵、价值损失、运行时间及收敛速度等多个指标，利用14节点NSFNET网络拓扑仿真实验。结果表明:当信道中包含18个波长时，与传统KSP-FF算法相比，所提出的路由波长分配算法的阻塞率降低了0.06，资源利用率提高了0.02，但运行时间有增加；在波长数超过45以后，与传统KSP-FF算法相比，所提算法保持阻塞率和资源利用率的同时，运行时间开始降低；当信道中包含波长数为58时，与传统KSP-FF算法相比，所提算法运行时间减少了0.07 ms。由此可见，提出的算法使路由选择和波长分配得到了优化。
|
| 关 键 词： | 光传送网 路由波长优化 强化学习 路由选择 波长分配 |
| 收稿时间： | 2022-02-07 |
Optimization of routing and wavelength optimization algorithm for optical transport network based on reinforcement learning |
| |
| Affiliation: | 1.Department of Electronic and Communication Engineering, North China Electric Power University, Baoding 071003, China2.Hebei Key Laboratory of Power Internet of Things Technology, North China Electric Power University, Baoding 071003, China |
| |
| Abstract: | Aiming at the routing and wavelength problems of dynamic services in optical transport network, a deep routing wavelength assignment algorithm based on reinforcement learning is proposed. The algorithm is based on a software defined network architecture, flexibly adjusts and controls the optical transport network through reinforcement learning, and realizes the optimization of the optical network routing wavelength assignment strategy. For the problem of routing selection, combined with the wavelength usage on the link, the A3C algorithm is used to select the appropriate route to minimize the blocking rate; for the problem of wavelength assignment, the first fit algorithm is used to select the wavelength. Considering multiple indicators such as blocking rate, resource utilization, policy entropy, value loss, execution time, and speed of algorithm convergence, the 14-node NSFNET network topology simulation experiment is implemented. The results show that when the channel contains 18 wavelengths, compared with the traditional KSP-FF algorithm, the blocking rate of this routing wavelength assignment algorithm is reduced by 0.06, and the resource utilization rate is increased by 0.02, but the execution time is increased. When the number of wavelengths exceeds 45, compared with KSP-FF, the proposed algorithm maintains the blocking rate and resource utilization, while the execution time begins to decrease. When the number of wavelengths is 58, compared with KSP-FF, the proposed algorithm's execution time is reduced by 0.07 ms. It can be seen that the proposed algorithm optimizes the routing and wavelength assignment. |
| |
| Keywords: | |
|
| 点击此处可从《红外与激光工程》浏览原始摘要信息 |
|
点击此处可从《红外与激光工程》下载全文 |
|
