A Unification of ML-Optimal Tree-Search Decoders

In this paper, a generalization of tree-search algorithms with maximum likelihood (ML) error rate performance for the detection in multiple-input multiple-output (MIMO) systems is presented. The Schnorr-Euchner sphere decoder, Djikstra's algorithm applied to MIMO detection, and an ML- optimal variant of the K-best detector are shown to be special cases of the presented unification. The complexity of the detection procedure is characterized by three different measures, relevant for VLSI implementations, and the corresponding trade-offs between hardware complexity and throughput are investigated. Finally, efficient detector configurations are identified which are suitable for MIMO detection of higher-order modulation alphabets and for large numbers of transmit antennas.