Research On Beam Forming Of MIMO Transmitter Based On Semideifnite Programming

Unlike the phased-array antennas, MIMO array allows independent signals to betransmitted at different antennas. Through this additional spatial diversity gain, MIMOarray can deliver higher spatial detection and filter capabilities. In this thesis, we focuson the beam forming design of MIMO transmitter. The design’s goals are toapproximate transmit beampattern to a desired one, and to minimize auto-correlationsidelobes at a given spatial angel and cross-correlation levels between different spatialangels through optimization of the transmitter’s constant modulus signals sequence.The main contributions of this thesis are as follows:1. The beam forming problem is modeled as a4-order polynomial problem withconstant modulus constraints of the transmitted signals. Because the original model isnonconvex, we relax it as a convex optimization model that can be solved usingpolynomial time through semidefinite relaxation and rank minimization procedures.And then we analyze this algorithm’s computation complexity. Simulation results showthat the convex optimization model can obtain the desired beampattern and lowauto-correlation sidelobes and cross-correlation levels.2. To reduce the computation complexity of convex optimization model, weconvert the original constrained problem to an unconstrained problem by exploiting thespecial structure of constant modulus constraints, and then propose a speedy algorithmbased on Quasi-Newton method with L-BFGS approach. Both the theoretical analysisand simulation results show that this new algorithm can not only obtain the desiredtransmitted signals, but also achieve a much lower computation complexity comparedwith the convex optimization method.