Research On Intelligent Routing Mechanism In Highly Dynamic Ad Hoc Networks

With the continuous development of wireless communication technology,mobile Ad Hoc network(MANET)is gradually emerging as a new technology.Mobile Ad Hoc network is a kind of wireless mobile communication network with no center,which can be built quickly without relying on various basic equipment.It is often used in military communication and personal communication.Due to the dynamic characteristics of MANET,the routing in MANET is very challenging.In many complex network environments,the traditional routing protocols in Ad Hoc networks cannot cope with the routing problem of differentiated service requirements,and the development of artificial intelligence technology provides a new idea to solve the routing problem in complex scenarios.Thesis is inspired from it,combines the relevant algorithms of reinforcement learning and graph neural network,studies the routing mechanism in different scenarios.The main research contents are as follows:First,thesis studies routing mechanism for load balancing in Ad Hoc networks.Due to the limited link capacity,with the increase of traffic flow in the network,the aggregation of traffic flows will lead to load imbalance and link congestion.Therefore,from the perspective of load balancing,thesis proposes a link overhead dynamic adjustment algorithm based on Deep Deterministic Policy Gradient(DDPG)algorithm,which realizes the dynamic optimization and adjustment of route through the intelligent dynamic decision of link routing overhead.Numerical results show that compared with the traditional shortest path algorithm and heuristic algorithm,this algorithm helps to balance the network load,improve the success rate of traffic flow transmission and increase the network throughput.Then we study the routing mechanism of UAV network based on energy optimization.In the UAV network,the energy of nodes is limited.The depletion of node energy will cause node failure,network partition and even paralysis.Therefore,thesis hopes to optimize the node energy.According to the characteristics of UAV network without center,the Graph Attention Network model(GAT)is introduced to learn the correlation between different node characteristics,so as to realize local information interaction.In order to adapt to the UAV distributed network,thesis proposes a distributed Actor Critic(AC)algorithm and designs the intelligent dynamic adjustment algorithm of UAV network link weight.Through simulation,compared with the traditional shortest path algorithm and heuristic algorithm,our proposed algorithm can reduce energy consumption and improve the network life.Finally,thesis studies the intelligent routing mechanism based on alternative path in Ad Hoc networks.Traditional routing protocols only produce a single path.In the actual transmission process,with the increase of traffic in the network,it is easy to lead to excessive concentration of traffic flow,resulting in network congestion and affecting network transmission performance.Thesis proposes an alternative path selection algorithm based on deep Q network(DQN).The algorithm can dynamically perceive the changes of the environment,learn and predict the traffic characteristics and the state of the path in the network,and ensure the transmission performance of the network by adaptively adjusting the route.The simulation results show that compared with the traditional shortest path algorithms,random algorithm and greedy algorithm,our proposed algorithm demonstrates significant performance gain in improving the success rate of traffic flow transmission and increasing the network throughput,and it also has the function of load balancing.