基于Q学习的无人机自组网智能过滤路由算法OA

Due to its dynamic topology and high mobility,flying ad hoc networks face challenges such as high routing time delay,unstable links and routing holes.Conventional routing protocols exhibit limited adaptability in complex dynamic environments,whereas reinforcement learning methodologies present innovative solutions to these issues.To solve these problems,an intelligent filtering routing algorithm based on Q-learning is proposed.UAV nodes perceive neighbor states by broadcasting Hello messages regularly,and construct state space based on link stability and neighbor topology information,employ a three-dimensional convex hull filtering mechanism to reduce the state space scale and mitigate Q-learning computational overhead.The multi-objective reward function of the algorithm com-prehensively considers the single-hop delay and link stability design to achieve multi-objective joint optimization.At the same time,the learning rate and discount factor are adaptively adjusted by combining topological information to enhance the environmental adaptability of the routing strategy.In addition,a HELLO message pre-learning mechanism is introduced,leveraging the maximum Q-values carried by neighboring nodes to accelerate Q-table convergence.Simulation results demonstrate that the proposed algorithm outperforms existing protocols such as QRF,QGeo,and GPSR in terms of packet delivery ratio and end-to-end delay,effectively improving routing performance in dynamically changing topological environments of flying ad-hoc networks.