移动边缘网络中基于双深度Q学习的高能效资源分配方法

Energy-efficient resource allocation method in mobile edge network based on double deep Q-learning

为了提升移动边缘网络中系统的能量使用效率,面向多任务、多终端设备、多边缘网关、多边缘服务器共存网络架构的下行通信过程,提出了一种基于双深度Q学习(DDQL)的通信、计算、存储融合资源分配方法。以任务平均能耗最小化为优化目标,设置任务时延和通信、计算、存储资源限制等约束条件,构建了对应的资源分配模型。依据模型特征,基于DDQL框架,提出了适用于通信和计算资源智能决策、存储资源按需分配的资源分配模型和算法。仿真结果表明,所提出的基于DDQL资源分配方法可以有效地解决多任务资源分配问题,具有较好的收敛性和较低的时间复杂度,在保障业务服务质量的同时,相对于基于随机算法、贪心算法、粒子群优化算法、深度Q学习等方法,降低了至少5%的任务平均能耗。

To improve the system energy efficiency in mobile edge networks,a resource allocation method based on double deep Q-learning(DDQL)for integration of communication,computing,storage resources was proposed for the downlink communication process under the network architecture of multiple tasks,end devices,edge gateways and edge servers.A resource allocation model was constructed,which took the minimization of average energy consumption of tasks as the optimization goal and set the constraints of task delay limits and communication,computing,and storage resource limits.According to the model characteristics,a suitable resource allocation model and method based on DDQL framework was proposed to make intelligent allocation decisions for communication and computing resources and allocate storage resources on demand.Simulation results show that the proposed DDQL-based solution can effectively solve the multi-task resource allocation problem with good converge and low time complexity,and it reduces the average energy consumption of tasks by at least 5%compared with the solving methods based on random algorithm,greedy algorithm,particle swarm optimization algorithm and deep Q-learning while ensuring the quality of service.