| 摘要: |
| 城市三维车载自组网（Three-Dimensional Vehicular Ad-hoc Network，3D-VANET）中往往存在多种道路形式，每种道路网络拓扑变化特点各异，平面路由协议不能根据道路特点动态调整选路策略，不宜直接用于3D-VANET，为此设计了一种基于模糊逻辑和Q学习的拓扑感知路由协议。该协议通过模糊逻辑方法感知网络拓扑变化与网络负载情况动态调整信标间隔，以平衡邻节点信息准确性与控制开销成本。在此基础上，采用Q学习算法对网络建模，根据链路质量以及链路质量变化调整Q学习算法参数，以灵活选择下一跳转发节点，更好适应网络拓扑的频繁变化。仿真结果表明，与对比协议相比，该协议有利于降低控制开销，同时提高包投递率和减少平均端到端时延。 |
| 关键词: 三维车载自组网（3D-VANET） 路由协议 拓扑感知 模糊逻辑 Q学习 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目（61371091，61801074)；大连市科技创新基金（2019J11CY015） |
|
| Topology aware routing protocol based on fuzzy logic and Q-learning for 3D-VANET |
| ZHENG Zetao,HE Rongxi,ZHOU Yu |
| (Information Science and Technology College,Dalian Maritime University,Dalian 116026,China;Information Science and Technology College,Dalian Maritime University,Dalian 116027,China;Information Science and Technology College,Dalian Maritime University,Dalian 116028,China) |
| Abstract: |
| In urban three-dimensional vehicular ad-hoc network(3D-VANET),there are often many kinds of roads.The topology change of different roads has different characteristic.Since plane-based routing protocols cannot adjust routing strategy according to the road characteristic,it cannot be directly utilized in the 3D-VANET.Therefore,a topology aware routing protocol based on fuzzy logic and Q-learning(TARP-FQ) is proposed.In the TARP-FQ,topological change and network load are acquired though fuzzy logic to dynamically adjust the beacon interval for balancing the network overhead and the neighbor information accuracy.Moreover,the Q-learning algorithm is employed to model the network,and its parameters are adjusted according to the link quality and link quality changes for flexible selection of next forwarder,which can better adapt to the frequent topology change.The simulation results show that TARP-FQ is beneficial to improve packet delivery ratio and reduce average end-to-end delay and control overhead compared with the benchmark routing protocols. |
| Key words: three-dimensional vehicular ad-hoc network(3D-VANET) routing protocol topology aware fuzzy logic Q-learning |
