Capacity Analysis And Detection Algorithm Research For Massive Mimo In D2D Scenario

With the emergence of new mobile application scenarios and demands,mobile communication techinologies are facing the chanllenge brought by large amount of data transmission and device connection.Massive multiple-input-multipl-output(massive MIMO)and device to device communications(D2D)are two key technologies on which the fifth generation mobile communication system relies to support the scenario of enhanced mobile broadband and masive machine-type communications.The high efficient fusion of massive MIMO and D2D,especially leveraging the advantage of massive MIMO to mitigate D2D interference,is with imporatant research value.Aiming this problem,this dissertation studies the ergodic capacity and signal processing about using massive MIMO in base station's transmitter,as well as the high efficient detection algorithms for inerference mitigation when massive MIMO is used in receiver.At first,this dissertation studies the ergodic acheivable rate about using massive MIMO with antenna selectoin to mitigate the interference between downlink and D2D link.After derived the close-form formula,the upper and lower bound of the ergodic acheivable rate is obtained with Jensen inequality.The system's behavior under noise-limited scenario and interference-limited scenario is analyzed with the pair of bounds.Based on these results,this dissertation analyzed the impact of antenna number to the ergodic rate.Theoretical analysis and simulation shows that the ergodic rate of antenna selection is saturated with antenna number increasing.Furtherly,beamforming and interference zero-forcing were studied and proved to have logarithmic ergodic rate with antenna number.Secondly,for the scenario with spatial correlated channel where the number of D2D interferors is large and interference can be equivalent as white noise,this dissertation proposed two novel algorithms:lattice reduction aided belief propagation and lattice reduction aidied expectation propagation for massive MIMO detection.Based on the analysis about spatial correlation's impact on belief propagation,the MMSE aided constellation determination and symbol probability transforming using Gaussian approximation are proposed to realize the convergence of lattice reduction and belief propagation.Same method is also used to merge lattice reduction and expectation propgation.The proposed algorithms lead to 3?7dB gain and can be used to mitigate uplink-D2D interference effectively.Thirdly,for the scenario with spatial correlated channel where the number of D2D interferors is small and interference cannot be equivalent as white noise,this dissertation proposed lattice reduction aided whiten belief propagation and lattice reduction aidied whiten expectation propagation for massive MIMO detection.The interference whitening scheme was jointly used with lattice reduction,belief propagation and expectation propagation.The proposed algorithms leads to 5?10dB gain in the target scenario and can be used to mitigate uplink-D2D interference effectively.Finally,a massive MIMO prototyping based on general purpose x86 server was established.The solution based on FPGA isolation between RF unit and general-purpose IT platform was proposed to guarantee real-time performance.Also the parallel processing upon server farm,multi-cores,and data set are utilized for accelerating baseband signal processing.Over-the-air test upon it verified the effectiveness of lattice reduction aided expectation propagation algortihm.This prototyping system accelerated the verification of algorithms and explored the flexible baseband signal processing.