Efficient Stochastic LASF codes for MIMO-OFDM systems

High rate MIMO-OFDM systems can be adapted to improve symbol error rate (SER) by employing effective Lattice Aided Space Frequency (LASF) codes. Such high rate systems boasts high parametric dimensions to an extent that analytical formulation becomes restrictive, and in most cases impossible. The design of LASF codes have recently been studied under gradient-based stochastic methods. In this paper a low latency and low complexity novel stochastic method to design LASF codes is proposed. This is achieved by simultaneously perturbing the LASF generator matrix G and, from the noisy empirical measurements of the objective function, estimating the gradient. Algorithm convergence, in terms of the number of iterations, is faster than any other known stochastic methods. In fact, it is five times faster than the gradient based schemes. Comparing with an average randomly generated LASF matrix G,, it is demonstrated that in quasi-static frequency selective channels the SER is observed to halve at a SNR of 30dB. While having significantly lower computational complexity, simulations confirm that the novel stochastic method achieves improved SER performance across the entire SNR range. Furthermore, latency is significantly reduced since each computation requires a single pair of noisy measurements of the SER per iteration irrespective of the dimension of the system.