Abstract
The tree representation of the MIMO detection problem is illuminating for the development, interpretation, and classification of various detection methods. One method, based on the Dijkstra's search algorithm, pursues tree exploration according to a bounded, sorted list of tree nodes. Since sorting directly affects tree exploration and truncation in this tree-search method, it is critical to the performance. Motivated by the observation that sorting according to nodes' path metric, as in the conventional algorithm, does not adequately represent the "goodness" of nodes, a new probabilistic sorting rule is developed by innovatively exploiting the statistical properties of the path metric to yield more effective sorting. The relationship between the probabilistic sorting and the conventional one is established, and new features of the probabilistic sorting are presented. The effectiveness of the proposed method is demonstrated by computer simulation, where the new method outperforms the previous tree-search method in achieving near-ML detection performance, and meanwhile offers significant complexity reduction compared to the previous tree-search method. © 2011 IEEE.