Active User Detection And Channel Estimation In Massive Grant-free Multiple Access

The popularization of terminal equipments has led to the rapid development of the mobile Internet and the Internet of Things.The mobile communication network not only needs to satisfy the requirements of communication among people,but also needs to satisfy the requirements of communication among machines.One of the important scenarios in the fifth generation(5G)mobile communication system is machine to machine(M2M).M2M can provide communication services for a large number of devices.These devices have low complexity and low power consumption.The transmitted data by devices has characteristics of low speed and small data packets.The current long term evolution(LTE)system is not suitable for M2M.When facing a large number of users' connections,the contention-based multiple access in LTE will cause huge access signaling overhead.The grant-free multiple access can avoid access signaling overhead and reduce transmission delay and terminal power consumption.In addition,massive multiple input multiple-output(MIMO)technology can increase system throughput,improve transmission reliability and improve spectrum efficiency and energy efficiency.Massive MIMO is considered as one of the core technologies in the future mobile communication system.This thesis focuses on active user detection and channel estimation for the grant-free multiple access in M2M.First of all,this thesis compares the contention-based multiple access and the grant-free multiple access.User terminals in uplink transmission of current LTE require grants to acquire dynamic resource,which will generate huge signaling overhead.The grant-free multiple access eliminates the dynamic resource allocation.Once a user register on the network,pilot and time-frequency resource are allocated to the user.In the grant-free multiple access,when a user needs to transmit data,it can send data directly and do not need to re-establish the connection with the base station.Because of the lack of dynamic resource allocation,the base station can not know which users send data,that is,it is unknown which users are active.In order to recover the transmitted data,active user detection and channel estimation must be performed firstly.In this thesis,active user detection and channel estimation in grant-free multiple access are transformed into single measurement vector(SMV)and multiple measurement vectors(MMV).The pilot design in grant-free multiple access is transformed into the measurement matrix design.Secondly,considering a large number of users access to the base station,algorithms based on MMV for active user detection can not detect enough active users.In addition,these algorithms do not make full use of massive antennas at the base station.Using compressive sensing and Khatri-Rao product,active user detection is transformed from MMV to SMV An orthogonal matching pursuit based on Khatri-Rao product(OMP-KR)algorithm is proposed in this thesis.In addition,a scheme of pilot design based on ZC sequence is proposed.The iterative termination condition of OMP-KR algorithm is proposed when the base station do not know the number of active users.Simulation results demonstrate that compared with traditional MMV algorithms like simultaneous orthogonal matching pursuit(SOMP)and M-FOCUSS,OMP-KR can detect more active users.OMP-KR outperforms the traditional MMV algorithms in active user detection.When the number of antennas increases,the performance of OMP-KR will be better.Finally,the beam domain channel has sparsity in massive MIMO system.The number of active users which can be detected will increase and the performance of channel estimation will be better when utilizing the sparsity of the beam domain channel.In this thesis,an modified multi-orthogonal matching pursuit(MM-OMP)algorithm in beam domain is proposed.MM-OMP converts MMV into multiple SMVs.The channel estimation accuracy in each SMV will be improved by utilizing the sparsity of the beam domain channel.Then according to the joint sparsity of the channel matrix,active user detection and channel estimation can be achieved.Simulation results demonstrate that MM-OMP can achieve better performance of active user detection and channel estimation than traditional MMV algorithms like SOMP and M-FOCUSS.When the number of antennas increases,the performance of MM-OMP will be better.