Radius-adaptive sphere decoding via probabilistic tree pruning

In this paper, we propose a radius-adaptive sphere decoding algorithm that reduces the number of operations in sphere-constrained search while achieving performance close to ML decoding. Specifically, by adding a probabilistic noise constraint on top of sphere constraint, a more stringent necessary condition is provided, particularly at an early stage, and hence many branches that are unlikely to be selected are removed in the early stage of sphere search. From the simulation in a frequency selective channels with pruning probability ε = 0.03, it is shown that the computational complexity of proposed strategy reduces significantly (30-76%) over the original algorithm with negligible performance loss.