Research On Detection Of Active Users For Uplink Grant-Free Transmission Of Satellite Internet Of Things

With the development of satellite communication technology,the satellite Internet of Things system has become a reliable choice for Internet of things terminals in remote areas to access network services due to its strong resilience and wide coverage.It can realize the true "Internet of everything".Due to the wide coverage of satellite and the increasing number of terminals,the network will be faced with the challenge of random robust access.In scenarios with a large number of connections and light control,such as the Internet of Things,the wireless resources of the system are very limited.As a result,terminals compete for resources when accessing upstream,leading to data packet collision between users.At the same time,in order to reduce signaling overhead,reduce access delay and improve communication system efficiency,grant-free random access technology has become a research hotspot as a solution to support large-scale transmission of satellite Internet of Things.Grant-free transmission technology does not go through the connection establishment process,which meets the requirements of low power consumption and light control of the satellite Internet of Things.However,this will lead to the user’s data transmission behavior is often unpredictable at the satellite receiver,and the system cannot know which users are active,which brings difficulties to subsequent multi-user detection.Therefore,active user detection should be carried out first.At present,active user detection is usually performed in the code domain.However,the density of Internet of Things terminals is far higher than that of mobile communication users.Relying only on code domain identification will lead to a significant increase in address code dimensions and reduce frame efficiency.Moreover,many collection and monitoring Internet of Things terminals have weak capabilities and cannot report their own positions,leading to the loss of this one-dimensional access auxiliary information,which is not conducive to the improvement of detection performance of active users.Aiming at the development trend of space-borne phased array in the future,this thesis introduces the idea of spatial identification and adopts the multi-domain joint method to detect active users.The details are as follows:(1)Based on the processing advantages of future space-borne phased array antennas and the characteristics of distributed deployment of Internet of Things terminals,a cascade of code domain and spatial-domain active user detection method is proposed.The satellite receiver first identifies the user in the code domain for the received signal.In the process of algorithm iteration,if there are users using the same codeword,the direction of arrival is estimated,and the spatial information of users is utilized to further distinguish different users.The final active user estimation set is obtained through the cascaded code domain and spatial domain method.The simulation results indicate that compared with the traditional code domain detection scheme,the proposed method can achieve the purpose of user identification and improve the detection performance of active users in the case of codeword conflict.(2)In order to solve the problem of high complexity and poor performance under low signal-tonoise ratio conditions of the code domain and spatial domain concatenation method,a multidimensional joint active user detection scheme is proposed.According to the codeword sequence and azimuthal information of the signal,the joint codeword-angle codebook is created.Combined with the user’s codeword sequence and azimuthal information,the sparse reconstruction method was used to estimate the active users directly to achieve the effect of joint detection.The simulation results show that the proposed method can improve the detection success rate of active users when the signalto-noise ratio is low compared with the previous cascade scheme under the condition of no codeword conflict,and the algorithm is still effective to a certain extent when users have codeword conflict.(3)Aiming at the issue of limited identification performance of active user detection in compressed sensing algorithms when there are a large number of active users and high sequence correlation,the algorithm for active user detection based on machine learning is analyzed.According to the sparsity characteristics of users,a data set combined with code domain and spatial domain is created to learn active user characteristics,and a deep neural network model is constructed.The complex mapping between the received collision signal and the active user in the transmitted signal is learned,and the trained deep neural network model is used to detect the active users.The simulation results show that the proposed method can achieve active user identification when the sequence correlation is too high after learning the code and spatial features of the signal,and the detection success rate is improved compared with the detection scheme based on compressed sensing technology.