低时延网络多路通信数据无冲突调度仿真
Simulation of Collision Free Scheduling for Multi-Channel Communication Data in Low Delay Networks
郝聪妙 1孟晓丽 2王辉3
作者信息
- 1. 郑州工业应用技术学院计算机学院,河南 郑州 450000
- 2. 西安外事学院工学院,陕西 西安 710077
- 3. 西安工业大学计算机科学与工程学院,陕西 西安 710021
- 折叠
摘要
随着信息传播数据量的飞速增长,以及低时延网络的广泛应用,低时延网络无冲突调度的需求也进一步增加.为解决当前通信调度算法时效性差、调度效果不佳等问题,提出了基于强化学习演员-评论家算法的无冲突调度模型.模型首先对低时延网络架构进行分析,并对状态空间中的多路通信网络进行数据监测;然后基于受限制的马尔科夫决策对其进行无冲突调度,并基于贝尔曼方程计算结果进行策略调度;最后使用强化学习中的Actor-Critic算法对调度算法进行优化,通过Actor提出调度方案,通过Critic评价调度情况,进而达到无冲突调度的效果.实验结果表明,所提算法将冲突率降低了8.72％,将吞吐率提升了 7.58％,降低了数据在传输过程中丢失和乱码的概率,提高了通信数据的传输效率.
Abstract
With the rapid growth of information dissemination data and the widespread application of low-latency networks,the demand for conflict-free scheduling for low-latency networks has further increased.In order to address the issues of poor effectiveness and scheduling effectiveness of current communication scheduling algorithms,this paper proposes a conflict-free scheduling model based on the Actor-Critic algorithm in reinforcement learning.This model first analyzes the low-latency network architecture and monitors the data of multi-channel communication net-works in the state space.Then,based on the constrained Markov decision,conflict-free scheduling is performed,and the scheduling strategy is calculated based on the Bellman equation.Finally,the Actor-Critical algorithm in reinforce-ment learning is used to optimize the scheduling algorithm.The Actor is used to propose a scheduling plan,and Criti-cal is used to evaluate the scheduling situation,thereby achieving the effect of conflict-free scheduling.The experi-mental results show that the algorithm proposed in this article reduces the conflict rate by 8.72％,increases the throughput rate by 7.58％,reduces the probability of data loss and garbled code during transmission,and improves the transmission efficiency of communication data.
关键词
无冲突调度/低时延网络/强化学习/多路通信数据Key words
Conflict-free scheduling/Low latency network/Reinforcement learning/Multi-channel communication data引用本文复制引用
基金项目
教育部产学合作协同育人项目(221005812103741)
出版年
2024
NETL
NSTL
万方数据