基于深度强化学习的边云协同任务卸载策略

Edge cloud collaborative task offloading strategy based on deep reinforcement learning

为降低应用程序执行的时延和能耗,针对移动边缘计算环境,提出一种边云协同场景下基于深度强化学习的任务卸载策略。通过建立边云协同架构下的网络模型、通信模型及计算模型,以最小化时延和能耗为系统目标,设计基于深度强化学习的DQN卸载策略,将每个用户产生的任务独立高效地放置在本地、服务器或者云端进行计算,并将计算结果与其它方法进行比较。实验结果表明,相较其它基线算法,该方法能更有效减少任务执行的开销,得到更优的卸载策略。

To reduce the delay and energy consumption of application execution,a task offloading strategy based on deep reinforcement learning in edge-cloud collaboration scenario was proposed for mobile edge computing environment.The network model,communication model and calculation model under the edge-cloud collaborative architecture were established,and taking the minimization of delay and energy consumption as the system goal,the DQN offloading strategy based on deep reinforcement learning was designed,the tasks generated by each user were placed independently and efficiently in the local,server or cloud for calculation,and the calculation results were compared with other methods.The results show that compared with other baseline algorithms,this method can more effectively reduce the cost of task execution and obtain a better offloading strategy.