This thesis proposes a modified LLL algorithm, which named fast convergence algorithm (FC), in lattice reduction (LR) aided detection which provides advantage of the convergence time compared to conventional LLL reduction algorithm. Generally, LLL alg...
This thesis proposes a modified LLL algorithm, which named fast convergence algorithm (FC), in lattice reduction (LR) aided detection which provides advantage of the convergence time compared to conventional LLL reduction algorithm. Generally, LLL algorithm transforms MIMO channel, regarded as basis vector, into orthogonal-like one by using iterative basis reduction which is mainly classified into three functional processes: a Gram-Schmidt orthogonalization, a size-reduction, and a basis swapping. The proposed FC algorithm uses the swapping condition that the all of size reduced basis vectors are sorted out in the descending order of their basis vectors at once while conventional LLL algorithm swap only adjacent basis vectors per one-iteration. Since the same time complexity per iteration is applied for both two algorithms, the convergence time depends on the number of iteration needed. Therefore the convergence time to terminate the algorithm can be decreased by iteration number differences between two algorithms. Moreover, this proposed FC algorithm also provides the shorter resultant basis vectors. Results for these properties are developed, and illustrated by analytical and numerical analyses.In addition, three different schemes to reduce the complexity of LR in specific condition are proposed in this thesis: One is sorted basis reduction method, second is consecutive lattice reduction (CLR), and third is pre-lattice reduction (PLR). First sorting out basis vectors at the initial point helps LR algorithm (whether LLL or FC) to achieve the reduced basis vectors with reduced number of iterations. Secondly, time-correlated nature of the channel can be used in reducing the iteration complexity of doing LR by taking advantage of the information of LR results in the previous symbol time. Finally, PLR applied in spatially correlated channel reduces the number of iteration of LR by pre-multiplying the channel matrix by transformation matrix which reduces the square root of the transmitter correlation matrix. These schemes may provide a better stating point for the lattice reduction algorithm.Simulation result shows that the proposed FC algorithm has fast convergence time and three different schemes have low computational efforts due to small number of iteration taken, which leads to fast convergence and reduced complexity for specific condition.