边缘计算场景下基于强化学习的应用最优部署

为有效解决城市范围内智能公共交通应用程序的布局问题,制定总代价最小化的应用布局优化策略MIN-COST,以降低应用程序部署的总代价为目标,同时满足应用程序服务延时要求。通过提出一个基于深度强化学习技术优化公交边缘应用程序部署的一般框架,可以从历史经验中学习到最优化部署方法,相对于一般启发式算法更加快速。将仿真结果与其它部署策略进行比较,验证了所提策略可以在保证服务时延的基础上有效降低应用程序服务总代价。

Optimal deployment of applications based on reinforcement learning in edge computing scenarios

To solve the layout problem of intelligent public transportation applications in the city effectively,an application layout optimization strategy named MIN-COST with a minimum total cost was developed.This strategy targeted the overall cost of reducing application deployment while meeting application service latency requirements.By proposing a general framework for optimizing the deployment of public transport edge applications based on deep reinforcement learning techniques,it was possible to learn from the historical experience to optimize deployment methods,which was faster than general heuristic algorithms.Comparing its simulation results with that of other deployment strategies,it is verifies that the proposed strategy can effectively reduce the total cost of application services based on the guaranteed service delay.