| As the key to post-disaster emergency rescue,emergency communication networks are often used for information interaction between command centers and disaster areas to improve the efficiency of rescue work.With the advantages of high mobility,all-terrain adaptability and flexible deployment,unmanned aerial vehicle(UAV)can easily access some disaster areas that are inaccessible to humans.By coordinating a large number of UAVs,Flying Ad Hoc Network(FANET)is built to achieve concurrent transmission of information after post-disaster with multi-hop link.However,due to the high-speed mobility of UAVs,rapid changes in network topology can cause failures in routing and resource scheduling schemes,preventing the provision of real-time services to disaster victims and rescuers.Therefore,to improve the reliability of FANET in post-disaster,this thesis combines graph theory and reinforcement learning to design stable routing and dynamic resource scheduling schemes in FANET,abstracting UAVs as points and routing or resource sharing relationships in the network as edges,incorporating multi-dimensionality and dynamism of FANET into the graph model,and obtaining efficient and fast solutions with the help of reinforcement learning.The main work and contributions of this thesis are as follows:For the case of routing failure caused by link failure or local node damage of FANET in the post-disaster,this thesis proposes a Q-learning-based stable routing algorithm,aiming to maximize the stability of the path.Firstly,the three-dimensional Gaussian Markov movement model of the UAV is improved by considering the mobility of the disaster victims so that the UAV can quickly collect emergency information from the disaster victims;Secondly,the link stability is designed according to the transmission mode,link quality and relative movement speed among the UAVs,and the spectrum resources occupied by each path are considered to establish the multipath stability optimization objective function.Then,each path in the network is described as a hyperedge in the hypergraph,and the hypergraph matching theory is used to transform the multi-path stability optimization problem into the problem of solving the maximum weight independent set in the weighted hypergraph,and the problem is solved by reinforcement learning.Finally,the simulation results show that the proposed algorithm can achieve fast selection of stable routes in FANET in post-disaster and can solve the problem of routing instability in high dynamic scenarios.Aiming at the channel conflict problem caused by unreasonable scheduling of FANET communication resources under the limited conditions of complex environment after disasters,this thesis proposes an adaptive multi-channel time division multiple access(TDMA)scheduling algorithm based on Q-learning with the objective of minimizing the number of network conflicts.Firstly,a dynamic topology model of FANET is established using graph theory methods in post-disaster,and the vertex interference graph is constructed through the interference relationship between UAVs,and the time slot and channel allocation problem is transformed into a dynamic double coloring problem based on graph coloring theory.Secondly,considering the high-speed mobility of UAVs,the dynamic double coloring problem of vertices is solved by using reinforcement learning theory,and the trade-off between the convergence speed of the algorithm and the optimal solution exploration capability is optimized by adaptively adjusting the learning parameters to respond to the dynamic changes of the graph.Finally,compared with existing algorithms,simulation results show that the adaptive multi-channel TDMA scheduling algorithm based on Q-learning-proposed in this thesis can achieve trade-off optimization between network communication conflicts and convergence speed,and can solve the problem of resource allocation decision and fastchanging topology adaptation in post-disaster highly dynamic scenarios.The thesis contains 23 figures,8 tables and 98 references. |
PDF Full Text Request