With the proportion of intelligent services in the industrial internet of things(IIoT)rising rapidly,its data dependency and decomposability increase the difficulty of scheduling computing resources.In this paper,we p...With the proportion of intelligent services in the industrial internet of things(IIoT)rising rapidly,its data dependency and decomposability increase the difficulty of scheduling computing resources.In this paper,we propose an intelligent service computing framework.In the framework,we take the long-term rewards of its important participants,edge service providers,as the optimization goal,which is related to service delay and computing cost.Considering the different update frequencies of data deployment and service offloading,double-timescale reinforcement learning is utilized in the framework.In the small-scale strategy,the frequent concurrency of services and the difference in service time lead to the fuzzy relationship between reward and action.To solve the fuzzy reward problem,a reward mapping-based reinforcement learning(RMRL)algorithm is proposed,which enables the agent to learn the relationship between reward and action more clearly.The large time scale strategy adopts the improved Monte Carlo tree search(MCTS)algorithm to improve the learning speed.The simulation results show that the strategy is superior to popular reinforcement learning algorithms such as double Q-learning(DDQN)and dueling Q-learning(dueling-DQN)in learning speed,and the reward is also increased by 14%.展开更多
基于物联网技术开发的智能家居系统解决了异构网络内家电的互联问题,而其管理系统仍面临着管理平台单一、可扩展性差以及较低的用户体验等问题。在利用ZigBee协议构建家庭无线传感器网络的基础上,提出一种基于RESTful Web Services的智...基于物联网技术开发的智能家居系统解决了异构网络内家电的互联问题,而其管理系统仍面临着管理平台单一、可扩展性差以及较低的用户体验等问题。在利用ZigBee协议构建家庭无线传感器网络的基础上,提出一种基于RESTful Web Services的智能家居管理系统设计方案,增强了系统的可扩展性和跨平台能力,所开发的基于Android的智能家居客户端有效提升了用户体验。展开更多
基金supported by the National Natural Science Foundation of China(No.62171051)。
文摘With the proportion of intelligent services in the industrial internet of things(IIoT)rising rapidly,its data dependency and decomposability increase the difficulty of scheduling computing resources.In this paper,we propose an intelligent service computing framework.In the framework,we take the long-term rewards of its important participants,edge service providers,as the optimization goal,which is related to service delay and computing cost.Considering the different update frequencies of data deployment and service offloading,double-timescale reinforcement learning is utilized in the framework.In the small-scale strategy,the frequent concurrency of services and the difference in service time lead to the fuzzy relationship between reward and action.To solve the fuzzy reward problem,a reward mapping-based reinforcement learning(RMRL)algorithm is proposed,which enables the agent to learn the relationship between reward and action more clearly.The large time scale strategy adopts the improved Monte Carlo tree search(MCTS)algorithm to improve the learning speed.The simulation results show that the strategy is superior to popular reinforcement learning algorithms such as double Q-learning(DDQN)and dueling Q-learning(dueling-DQN)in learning speed,and the reward is also increased by 14%.
文摘基于物联网技术开发的智能家居系统解决了异构网络内家电的互联问题,而其管理系统仍面临着管理平台单一、可扩展性差以及较低的用户体验等问题。在利用ZigBee协议构建家庭无线传感器网络的基础上,提出一种基于RESTful Web Services的智能家居管理系统设计方案,增强了系统的可扩展性和跨平台能力,所开发的基于Android的智能家居客户端有效提升了用户体验。
