Applications of SDR exact-ML criterion to tree-searching detection for MIMO systems

In previous work, we proposed the positive diagonal values (PDV) criterion, which is an exact-ML criterion of semidefinite relaxation (SDR) optimality condition. In this paper, we apply the PDV criterion to the tree-searching based MIMO detection by two ways. The first application is node-pruning algorithm for depth first search such as sphere decoding (SD). The proposed node-pruning algorithm using PDV criterion is not based on the Euclidean distance mostly used for node-pruning algorithm, instead, it uses an absolute test in each node so that it can be worked independently with many existing node-pruning algorithms. Furthermore, the proposed node-pruning algorithm can guarantee the exact-ML performance and reduce the number of nodes visited significantly. The second application is K-best algorithm of breadth first search. The proposed K-best algorithm takes K candidates at each stage based on PDV criterion. As a result, the proposed K-best algorithm can achieve near-ML performance.