Research On User Detection For Grant-Free Random Access Systems

The rapid development of wireless communications has promoted the prosperity of Internet of Things(IoT).With the rapid increase in the number of IoT devices,the subject of wireless traffic has gradually changed from human communication to massive machine-type communication(mMTC).In such cases,traditional grant-based access protocols are not suitable due to the excessive signaling overhead and high latency.Therefore,3 GPP has proposed the grant-free random access(GFRA)protocol to support mMTC and other IoT scenarios,which significantly reduces signaling overhead and latency by allowing devices to access the network without getting a grant from the base station(BS).At the receiver,due to the lack of a prior information,the BS performs blind detection to identify active devices and estimate channel state information(CSI).However,the grant-free access system is characterized by low transmission power,highly random transmission and fast-changing environment,which has greatly decreased the performance of traditional detection schemes.Hence,it is necessary to propose novel GFRA detection schemes to detect the access status(AS)and identify active devices precisely.In view of this,this dissertation focuses the AS detection and user detection,aiming to improve the robustness and accuracy of detection algorithms by exploiting novel robust signal features and comprehensively designing transmission models as well as detection schemes.In particular,we propose phase difference-based status detection schemes with strong robustness for GFRA systems with low power,the composite preamble transmission models and corresponding user detection schemes for GFRA systems with high randomness,and the power-split-based preamble transmission and detection schemes for GFRA users in cell-free networks.The main contributions of the dissertation are summarized as follows.First,the robust AS detection schemes based on the phase difference are proposed to reduce the influence of low power,unknown a priori information,uncertain noise variance and frequency offset on AS detection in GFRA systems.In particular,the distribution of phase differences is investigated based on a single-antenna GFRA system,and it is found that the phase difference is less affected by flat fading channels,and its probability density function(PDF)is almost unaffected by the noise variance and frequency offset.In view of these favorable properties,two AS detection algorithms,named phase difference distribution shape recognition detection algorithm(SRD)and phase difference distribution cosine feature extraction(CRD)are respectively proposed to achieve fast and robust AS detection against noise uncertainty and frequency offsets.Further,the above algorithms are extended to multi-antenna systems,where the AS detection algorithm based on the joint phase differenceeigenvalue difference feature is proposed.The joint feature is modeled and parameters are optimized to maximize the probability of detection.Simulation results show that the robustness and accuracy of the proposed schemes are significantly better than existing AS detection schemes.Second,based on the AS detection result,the BS performs user detection by processing preambles to identify active devices and estimate CSI.In this part,considering preamble collisions caused by the highly random selection of preambles,we propose the composite preamble-based user detection scheme to mitigate preamble collisions and thus improve user detection accuracy.In detail,utilizing phase-rotated non-orthogonal Zadoff-Chu sequences,a composite model containing orthogonal and nonorthogonal sequences is proposed,where orthogonal sequences facilitate accurate user detection and non-orthogonal sequences are used to expand the preamble pool.The theoretical analysis in terms of non-orthogonal interference and the dedicated detection scheme are provided.Numerical results show that using phase rotations can not only alleviate preamble collisions,but also reduce non-orthogonal interference,thereby reducing channel estimation error and improving user detection performance.Next,cyclic shifts are added to the above non-orthogonal sequences to further expand the preamble pool.By analyzing the effect of roots,cyclic shifts and phase rotations on the redesigned Zadoff-Chu sequences,an enhanced composite preamble based on root-shift-phase is proposed,which enlarge the preamble pool without adding extra interference.Simulation results verify that the proposed schemes outperform existing ones in terms of probability of detection,channel estimation error and success rate.Finally,the idea of improving user detection performance by comprehensively designing preambles and dedicated detection algorithms is applied to cell-free GFRA systems.To tackle the challenge caused by the lack of perfect uplink power control,the dissertation proposes an idea of power-splitting preamble transmission to improve user detection performance,where the effect of randomly distributed power is eliminated by power splitting ratios.Specifically,the same sequence is transmitted twice with different power-levels,so that the influence of unknown power can be eliminated by calculating the ratio of the two power levels.In addition,the ratio can also be used as an identification.In other words,devices choosing the same sequence can be distinguished by different ratios,so that colliding users can be detected through power-independent features.Then,according to the sequence and ratio detection results,uncertain power can be estimated and CSI estimation can be obtained accordingly.The power estimation error obeys a normal distribution with zero mean and rather low variance,which means that the channel estimation error is low under imperfect power control conditions.Simulations are performed to evaluate the proposed scheme.It is shown that with the same probability of preamble collisions,preambles based on power-splitting transmission can effectively improve the performance of active user detection and channel estimation.