Link abstraction models for multicarrier systems: A logistic regression approach

Abstract

Evaluation of complex wireless systems through simulations is commonly tackled with an abstraction of physical link behavior. This procedure is performed in 2 phases: Firstly, training is performed at the link level, resulting in a mapping function from signal-to-interference-plus-noise ratio (SINR) to block error rate (BLER); and secondly, physical layer signal processing is replaced by the resulting link-level mapping function. Several methods have been proposed in the literature for link abstraction. The well-known exponential effective SINR mapping (EESM) is likely to be the most extensively used procedure for multicarrier transmission. In short, EESM estimates an effective SINR from the whole set of SINRs of useful subcarriers by means of a single parameter mapping function, which must be previously optimized through training. Later on, BLER is assumed as that of the effective SINR over an additive white Gaussian noise channel. In this paper, we propose a novel method for BLER prediction named as logistic regression-based link method (LRLM). Mean and standard deviation of the SINR set can accurately capture BLER behavior for wideband multicarrier systems through a simpler function, thus reducing the amount of information to be stored compared to EESM. Moreover, LRLM is very flexible and may include extra predictors to reduce the number of different models to be stored. Performance results show that LRLM is able to predict accurately the BLER for all modulation and coding schemes and transport block sizes defined for Long-Term Evolution technology