Research On Multi-user Detection Algorithms Based On Angle Domain

Massive MIMO is one of the key technologies in 5G.Multi-user detection algorithm with high efficiency and low complexity is the core technology for the successful application of large-scale MIMO.Message passing algorithms,such as Belief Propagation(BP)algorithm and its various transformation improvement algorithms,have unique advantages in computational complexity or bit error rate performance.Aiming at the problem of multi-user detection in large-scale antenna array,the following research is conducted.In view of the high complexity of factor graph-based BP algorithm in signal detection in large-scale MIMO system and the problem of loop,a BP algorithm based on gaussian tree approximation is designed.In the Angle domain with channel sparsity,the algorithm complexity is reduced in the process of message passing by the gaussian approximation.For the problem that the BP algorithm cannot converge to the exact value in the cyclic factor graph,the factor graph can be transformed into a tree structure.Using Prim algorithm to find the optimal approximation tree and applying BP algorithm to the optimal approximation tree can solve the problem that the algorithm is difficult to converge because of the loop.Simulation results show that the Gaussian tree approximate BP detection algorithm based on the Angle domain can reduce the complexity of the algorithm and avoid the problem of accurate convergence due to the circulation problem.To solve the problem that multi-user detection in two-dimensional planar array has high computational complexity,GAMP algorithm is used in multi-user detection.The information in two-dimensional array has two Angle information,namely azimuth Angle and elevation Angle.Simplify the BP algorithm in message passing,and use AMP and GAMP algorithm for signal detection.AMP and GAMP algorithms avoid the complex probability density function transmitted by BP algorithm,and make full use of the mean and variance information in the message iteration.When solving the posterior probability density,it can efficiently reduce the computational complexity.Simulation results show that GAMP algorithm is close to BP algorithm in terms of bit error rate performance under the premise of small complexity.