Genetic designing of near-optimal training sequences for spatial multiplexing transmissions

This paper describes an efficient design method of training sequences for spatial multiplexing systems, where multiple antenna elements are utilized for highly spectrum-efficient communications. Recently, these wireless communication systems have been greatly interested in and investigated intensively, e.g. space-time coding (STC) scheme and Bell laboratories layered space-time architecture (BLAST). In these systems, the error-rate performance is seriously dependent on the residual channel estimation error. Therefore, good training sequences for channel estimations, that are short sequences and minimizing the estimation error, are desirable. However, it is not straightforward to construct the optimal sequences due to the large search space for multiple antenna systems. In this paper, we propose the design method using the genetic algorithms (GA) for optimizing training sequences, and demonstrate that the near-optimal sequences regarding least-squares (LS) criteria can be given efficiently. In addition, it is shown that optimized 16PSK sequences offer better performance than optimized BPSK sequences for short lengths.