Study On Relay Selection Algorithm Based On Multi-Armed Bandit In Underwater Acoustic Cooperative Communication Networks

As the our country for the marine resource strength is increasing,the requirement for underwater wireless communication technology has become more and more urgent.Due to the severely attenuation when electromagnetic wave propagation in the water,sound waves are the only form of energy known to mankind so far to achieve underwater long-range transmission.However,underwater acoustic channel is an extremely complex channel with time space frequency change,strong multipath interference,limited frequency band and high noise,at present it is the most difficult wireless communication channel.Cooperative communication technology can improve communication channel capacity,increase communication coverage and enhance transmission reliability.The same as radio frequency(RF)scenario,the relay selection problem(Relay Selection,RS)is the key problem need to be solved in underwater acoustic cooperative communication.However,due to the unique character of the underwater acoustic channel,the traditional relay selection strategy of the traditional RF cooperative communication can not be directly applied to the complex underwater acoustic communication environment.Therefore,this paper proposes a highly efficient,low cost,and highly applicable relay selection method for the complex underwater acoustic communication environment for users in different communication scenarios.This paper first analyzes the characteristics of underwater acoustic communication channel and the impact on the communication system.The sound speed characteristics in the ocean are introduced,and the road strength loss,shadow fading,multipath expansion,Doppler frequency shift and underwater acoustic channel noise are analyzed.Aiming at underwater acoustic cooperative communication relay technology,the cooperative relay selection principle of single source node and multi-source node is given respectively.By analyzing the particularity and complexity of underwater channel,it is proposed to model the relay selection problem in underwater sensor network as the decision problem of Multi-armed Bandit(MAB).Secondly,in the case of multiple relay of single source nodes,in view of the problem that the fixed probability distribution or attenuation model can not accurately describe the influence of channel environment on information transmission,combined with contextual bandit(CB)algorithm,a single source node underwater acoustic cooperative communication network relay selection model is set up,two relay selection algorithms of LinUCB and KernelUCB respectively based on the linear and nonlinear function relations between the channel environment information(temperature,salinity,relay node location and so on)and the network throughput,which can effectively improve the network throughput of the communication system and simplify the relay selection process.The simulation results show that the KernelUCB algorithm is more robust to the changing environment,and is more suitable for complex and changeable underwater communication environment than LinUCB algorithm.Finally,in the case of multiple source nodes and multiple relays,a distributed stable matching multi-user multi-armed bandit(DSMU-MAB)relay selection method is proposed for the imperfect channel state information(CSI)in dynamic underwater environment.The method adopts the stable matching theory and set the timer to realize the distributed requirement.It can not only effectively eliminate the user conflict but also avoid a large amount of information exchange.It reduces the computational complexity and solves the problem of the total throughput reduced due to the conflicts between users,thus improving the performance of the system.A distributed robust relay selection method(Distributed Stable Matching Multi-user Robust Multi-armed Bandit,DSMU-rMAB)is proposed,considering the influence of multi water environment factors(wind speed change,ship noise and so on)on relay selection.This method can effectively combat any changing communication environment,improve the robust robustness of the algorithm,and have strong applicability in the complex and changeable underwater communication environment.