Research On Key Technologies Of Machine Learning Networking For Large Scale And Fast Moving Robot Groups

With the development of mobile robot technology,mobile robot technology is widely used in various military and civil fields.The networking applications of mobile robot represented by UAV are developing towards the direction of large-scale and high-speed movement.Improving the quality of wireless communication is one of critical problems to be solved in large-scale fast mobile robot networking.Routing protocol and channel access algorithm have great impacts on the performance of wireless communication.Due to the large application scale and fast moving speed of UAV,the networks end-to-end hops are large and the contact time in the communication range is short.In view of the above problems,this thesis studies the large-scale UAV network routing protocol and fast mobile UAV-aided network channel access algorithm,in order to reduce the network end-to-end delay and improve the network throughput.The main work and innovations include the following two points:Firstly,in order to solve the problem of long end-to-end hops in large-scale UAV network,a fuzzy decision tree routing protocol(FDT)is proposed.In FDT,the number of hops is predicted by fuzzy decision tree,and the next hop node with the best comprehensive performance of delay and transmission success rate is selected from the minimum hop set of neighbors by exponential weighted average,so as to reduce the number of hops and ensure a higher transmission success rate.The simulation results show that the end-to-end delay of FDT is 72.1%,76.6%,74.5%,22.1%lower than that of RFLQGeo,MLProph,DOA and LBR,respectively.However,the transmission success rate of FDT needs to be improved.Secondly,in order to increase the communication contact time and solve the problem of low network throughput,a metric learning(ML)channel access algorithm is proposed.In ML,metric learning is used to optimize the flight trajectory of UAV base station to ensure that UAV base station has contact time with ground equipment enough to complete data collection in high-speed flight.Graph coloring method is used to reduce access conflict in channel access mechanism.The simulation results show that the throughput performance of ML is 38.18%,46.24%,49.15%and 26.76%higher than that of DLMA,iQRA,QLeaner and SLA,and the transmission success rate is also very good.However,the ML algorithm proposed in this thesis has limitations in delay and energy consumption.The research work of this thesis can provide reference for later scholars to study large-scale UAV network and fast mobile UAV-aided network.