Intelligent Topology Control Mechanism For UAV Group

In recent years,with the development and wide application of unmanned aerial vehicle(UAV)and the continuous rise of multi-UAV systems,UAV topology control technology has become a new research hotspot.The deployment of UAV networks is flexible and efficient,without the need to establish infrastructure in advance,and it is not limited to the network structure centered on the control base station.Its multi-hop communication mode breaks through UAV can only work within the base station communication range.It also effectively expands network coverage and has the potential to change the battlefield communication mode,but at the same time its multi-hop,self-organization,non-center,high mobile speed,and fast network topology changes also brings the design of UAV communication systems new challenges and requirements.This thesis combines bionic evolutionary algorithms and machine learning to study the topology control,clustering mechanism,and relay selection problems of UAV networks,and focuses on performance improvements such as network connectivity,energy consumption,and throughput.First,a distributed topology control scheme based on discrete particle swarm algorithm is designed for the scenario of large-scale unmanned aerial vehicle group cooperatively performing tasks in battlefield.At the beginning,the CAM algorithm is used to find the key nodes in the network.The network is segmented by these key nodes.Further,this thesis uses particle swarm algorithm to implement network topology control,and we abstract a mathematical model which seeks to minimize the energy consumption of the network.The optimization goal is to find the local degree-constrained minimum spanning tree.Effective topology control is implemented in each network area.Finally,the effectiveness of the proposed algorithm is verified by simulation experiments,and the network performance advantages are analyzed from the aspects of network connectivity,average node degree,average interference degree,average path length,average link length,and link robustness.Then,this thesis continues to study the clustering problem in UAV networks.In order to achieve the goal of fast clustering and effectively extend the network survival time in a highly dynamic UAV network scenario,we design a corresponding clustering mechanism.First,the optimal cluster number in the current network is calculated based on the equalized bandwidth,and then the k-means algorithm is used to quickly cluster the entire network.In order to ensure that the number of nodes in each cluster is roughly balanced,after k-means algorithm clustering is completed,a corresponding adjustment mechanism is also specially designed.Finally,the cluster head selection algorithm based on Deep Q-learning is implemented in each cluster.By sensing the changes in the surrounding environment in time,an appropriate reward and punishment mechanism is designed to make the long-term return of the entire network operation larger.Finally,simulation experiments were performed and compared with the traditional methods of highest node degree,weighted method,and random algorithm.We analyzed simulation results from the aspects of average link retention time,average cluster head time,and network energy consumption.The validity and correctness of the algorithm proposed in this thesis is verified.Finally,this thesis studies the problem of relay selection in UAV networks.The problem of relay selection is closely related to the transmission performance of the network.Excellent and efficient relay selection strategies can ensure the load balancing of each relay and the high throughput of the network.In the case of a large number of UAV nodes,the centralized solution will bring a lot of information overhead and calculations.Therefore,this thesis chooses a distributed Q-learning algorithm to solve this problem.Each UAV node determine its own strategy based on the surrounding environment and relay access information,so that the communication rate will meet the current task level requirements.Simulation analysis proves the effectiveness of the proposed algorithm.Compared with traditional greedy and random selection algorithms,this algorithm can achieve better network performance.In this thesis,we study the topology control,clustering mechanism,and relay selection strategy problems in UAV network in depth.Corresponding mathematical optimization models and Markov models are established.We design corresponding algorithms based on particle swarm algorithm,reinforcement learning and other theories,of learning and other theories,which provide some references and new ideas for the research of related problems of UAV networks in dynamic scenarios.