Edge Computing Dynamic Resource Allocation in Communication Network Delay Optimization
A deep reinforcement learning-based algorithm is proposed for the communication network delay optimization problem with dynamic resource allocation for edge computing.The algorithm establishes a mathematical model that comprehensively considers the constraints of heterogeneity,geographic dispersion,and energy causality of edge nodes,as well as multi-objective optimization metrics such as delay,energy consumption,and cost,to fully reflect the complexity and diversity of edge computing networks.The algorithm uses a multilayer perceptron as a policy function,and makes dynamic resource allocation decisions through the policy gradient method and stochastic gradient descent method to minimize the communication network delay.Simulation experimental results show that the algorithm can effectively reduce the communication network delay and improve the resource utilization efficiency and system performance under different scenarios and parameters,and has obvious advantages over fixed resource allocation algorithms,greedy resource allocation algorithms,and resource allocation algorithms based on genetic algorithms,especially in dealing with the dynamics and uncertainty of edge computing networks.The dynamic resource allocation algorithm based on deep reinforcement learning provides a new solution to the edge computing resource management problem,and also provides strong support for communication network delay optimization.
NETL
NSTL
万方数据
