Abstract
In this paper, we propose a computationally efficient iterative sequential detection algorithm that can achieve near-ML performance for V-BLAST systems. To reduce the receiver's efforts as much as possible while achieving the average bit error rate as small as possible, the proposed algorithm employs four essential ideas: i) adaptive number of iterative tree searching, ii) two early termination techniques, iii) branch metric calculation with limited constellation points, and iv) variable stack extension. More specifically, we first establish an M-ary tree structure using QR decomposition of the channel matrix. To make an iterative tree searching-friendly form, the full depth of this tree structure is then divided into the first tree depth and the remaining depths. Then, the algorithm generates an ordered list of candidate symbols at tree depth of one. It proceeds to iteratively search the remaining symbols followed by each candidate symbol at second-to-last depth, until it finds an optimal sequence. The proposed scheme features a very low receiver complexity and efficient storage usage. Its advantage stems from the well designed iterative tree searching based on the aforementioned ideas. Simulation results demonstrate that the proposed algorithm provides the near-ML performance with significant saving in complexity and storage.