| In the development of wireless communication technology,the Mobile Ad-hoc network(MANET)architecture is a widely used wireless communication network architecture.Based on its simple and fast networking characteristics,as well as its excellent flexibility and stability in the communication process,MANET has been widely used in disaster relief and military fields.As the application scope of MANET continues to expand,people have different requirements for the network in terms of delay,bandwidth,stability,and other aspects in different environments.However,in the wireless mobile scenario where the network structure changes rapidly,MANET faces many challenges in providing communication services that meet network service quality requirements.Based on such scenarios,this thesis studies intelligent routing technology in mobile ad-hoc networks from different perspectives.In the first part,we study the commonly used routing protocol in wireless ad-hoc networks,the OLSR routing protocol,and improved the MPR set selection algorithm in this protocol.The MPR set selection problem is an NP-complete problem,and it is difficult to obtain the optimal solution based on the greedy algorithm,which leads to redundancy in MPR node selection and uneven distribution of node load.In order to solve the problems of the original algorithm,this thesis proposes two MPR node selection algorithms based on genetic algorithm and ant colony algorithm,and designs a routing mechanism that can effectively reduce the additional flooding of TC messages.Simulation results show that compared with the greedy algorithm,the selection algorithm based on the evolutionary algorithm effectively reduces the redundancy in MPR selection and reduces the signaling overhead in the network.In the second part,we improve the traditional OLSR protocol routing method by changing the characteristic of OLSR routing only focusing on the minimum hop count path,and takes the business load of network nodes into the calculation of routing cost.A multipath routing algorithm is proposed to plan different data packets of the business flow to different paths for transmission,to avoid problems caused by load imbalance such as network congestion.We proposed two kind of multipath scheduling algorithm based on weighted round robin and deep reinforcement learning algorithm.Simulation result proves that the multi-path routing protocol based on deep reinforcement learning proposed in this thesis can effectively improve the network throughput and reduce packet loss caused by queuing delay in high business load scenarios.In the third part,another widely used routing protocol in wireless ad-hoc networks,AODV protocol is studied.The traditional AODV protocol that only targets the minimum hop count is difficult to ensure the stability of the selected link in high dynamic network scenarios,and also ignores the node’s load situation.The calculation of the optimal path considering factors such as delay,bandwidth,and load is a multi-constraint NP problem.Therefore,this thesis designs an improved AODV routing protocol based on particle swarm optimization to optimize path selection.Simulation shows that the improved protocol can find more stable and reliable routing paths in high dynamic scenarios.This thesis studies intelligent routing technology in mobile ad-hoc networks from different perspectives and proposes various routing optimization methods from multiple angles,effectively improving the performance and stability of the network. |
