基于强化学习的IEEE802.15.4网络区分服务策略  

IEEE 802.15.4 differentiated service strategy based on reinforcement-learning

作　　者：钱亮[1] 钱志鸿[1] 李天平[1] 全薇[1] 

机构地区：[1]吉林大学通信工程学院,吉林长春130012

出　　处：《通信学报》2015年第8期171-181,共11页Journal on Communications

基　　金：国家自然科学基金资助项目(61071073;61371092)~~

摘　　要：为了弥补IEEE 802.15.4协议原有区分服务机制的不足,提出了一种基于BCS(backoff counter scheme)与强化学习的区分服务策略。从终端节点出发,在原优先级区分服务策略的基础上增加BCS退避策略以解决流量较大场合业务区分问题;针对协调器节点,提出了基于强化学习的占空比调整策略,该策略能根据不同应用需求和环境变化自适应调整占空比。仿真结果表明,提出算法能针对不同环境满足高优先级业务性能需求,并能根据流量变化进行占空比调整,具有极强环境适应性。To provide better support in differentiated service for IEEE 802.15.4, a novel differentiated service mechanism was proposed based on BCS（back off counter scheme） and reinforcement learning. In terms of end-device, BCS backoff strategy was added to original priority-based differentiated strategy to solve the service differentiation problem under higher traffic condition. While for the coordinator, a reinforcement learning based duty-cycle adjustment algorithm was proposed to ＂self-learning＂ an optimal duty-cycle according to different application requirements and environmental changes. Simulation shows that the proposed algorithm can meet the performance requirements of high-priority service under different environments and adjust the duty-cycle when traffic is changed, which showed a strong environmental adaptability.

关 键 词：IEEE 802.15.4/LR-WPAN 区分服务 退避机制 强化学习 占空比 

分 类 号：TN92[电子电信—通信与信息系统]