Blanquez-Casado et al. EURASIP Journal on Wireless Communications and Networking (2016) 2016:20 DOI 10.1186/s13638-016-0518-3 RESEARCH Open Access eOLLA: an enhanced outer loop link adaptation for cellular networks Francisco Blanquez-Casado*, Gerardo Gomez, Maria del Carmen Aguayo-Torres and Jose Tomas Entrambasaguas Abstract Link adaptation (LA) process is a core feature for the downlink of 3GPP long-term evolution (LTE) and LTE-advanced (LTE-A). Through a channel quality indicator (CQI), the receiver suggests to the base station (BS) an appropriate modulation and coding scheme (MCS) according to the current channel conditions. In order to overcome any non-ideality in this process, the outer loop link adaptation (OLLA) algorithm is used to adaptively modify the mapping from signal-to-noise ratio (SNR) to CQI. OLLA basically modifies the measured SNR by an offset, according to whether data packets are received correctly or not, in order to adjust the average block error rate (aBLER) to a target. Although the OLLA technique has been extensively used, there exists a lack of analysis in the literature about its dynamics and convergence conditions. In this paper, a deep analysis of this algorithm has been carried out in order to cover this gap. From this analysis, we propose a new approach to the OLLA, the enhanced OLLA (eOLLA), which is able to adaptively modify its step size as well as to update its offset according to the reception conditions even if no data packets have been received. Thus, for LTE- and LTE-A-realistic scenarios, simulation results show that the proposed eOLLA outperforms the traditional OLLA, achieving a performance gain of up to a 15 % in terms of throughput. Keywords: Link adaptation, AMC, LTE-A, OLLA, BLER 1 Introduction A static selection of the values for the AMC thresholds The adaptive modulation and coding (AMC) process car- does not perform well in practical implementations as link ried out in the link adaptation (LA) is a crucial part of conditions are inherently variant. It is usual to adjust these current wireless communication systems. This technique thresholds by means of the well-known outer loop link allows to increase the data rate that can be reliably trans- adaptation (OLLA) technique, which was first proposed mitted [1] and has been adopted as a core feature in in [3]. Basically, OLLA modifies the SNR thresholds by an cellular standards such as long-term evolution (LTE) and offset [4, 5] which can be positive (making the MCS selec- LTE-advanced (LTE-A) [2]. tion more robust) or negative (when the CQI selection In the LTE and LTE-A downlink AMC procedure [2], was too strict). This offset is continuously updated based the user equipment (UE) has to suggest to the base sta- on the reliability of the received data blocks so that the tion (BS) an appropriate modulation and coding scheme average BLER is kept as close as possible to a predefined (MCS) to be used in the next transmission in order to target. keep the block error rate (BLER) below a target. The pro- Although there are works devoted to OLLA in the lite- posed MCS is signaled from the UE by means of a channel rature [6, 7], they typically address its performance from quality indicator (CQI). Typically, each CQI is associ- simulations, and the lack of a comprehensive analysis of ated with a particular signal-to-noise ratio (SNR) inter- its behavior in the literature is noticeable. Furthermore, to val; hence, MCSs are selected by mapping the estimated the best of our knowledge, previous works do not analyze instantaneous SNR into its corresponding SNR interval, the conditions under which the OLLA technique works defined by an upper and a lower threshold. properly. The first aim of this work is to cover this gap by carrying out a deep study of the OLLA technique. *Correspondence: [email protected] Department of Communications Engineering, Universidad de Málaga, Malaga, From this study, improvements in the implementation Spain of the traditional OLLA can be inferred. Thus, in this © 2016 Blanquez-Casado et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Blanquez-Casado et al. EURASIP Journal on Wireless Communications and Networking (2016) 2016:20 Page 2 of 16 paper, a different approach to the OLLA technique is pro- where A (, 0) is the probability of not being in outage, posed, the enhanced OLLA (eOLLA), which can signifi- that is, ∞ cantly improve the performance of the traditional OLLA. 1 −γ − ( ) = / = 0/ This paper is organized as follows. In Section 2, the A 0 e dγ e .(4) γ = AMC model for LTE used in this work is described, and 0 then a detailed description of the OLLA is carried out, According to Eq. (1), in order to evaluate the aBLER, including a study of its convergence conditions and its per- it is necessary to have at our disposal an expression for AWGN formance. In Section 3, the proposed eOLLA is presented. iBLERi , but to the best of our knowledge, this is Finally, Section 4 shows a comparison between both the not available in the literature when turbo coding is used. traditional OLLA and the proposed eOLLA in realistic Moreover, the exact value of the iBLER strongly depends scenarios based on the downlink of LTE and LTE-A, and on the specific decoder implementation [8]. Nevertheless, some concluding remarks are given in Section 5. since the iBLER metric represents the probability of being in one of two states {error, no − error},weproposetheuse 2 Outer loop link adaptation (OLLA) of binary logistic regression [9]. This regression is a binary To perform the AMC [1], the instantaneous SNR γ is esti- classifier based on one or more input variables. Thus, it mated at the UE to determine the current fading region i is a useful tool to model iBLER curves for each MCS i γ and, consequently, the transmission rate Ri (bits/symbol). over AWGN channels, for a given instantaneous SNR , At the UE, this instantaneous SNR is mapped into a by means of binary logistic functions as: certain CQI value, which is fed back to the BS. AWGN (γ ) ≈ (γ ) = 1 The set of SNR thresholds { } = defines the inter- iBLERi fi −α γ −α ,(5) i i 0,1,..,n 1 + e i0 i1 vals to map the estimated instantaneous SNR into its where the αi and αi values (see Table 1) are to be corresponding CQI, with 0 representing the minimum 0 1 found from the logistic regression over results of the required SNR for transmission (outage condition) and actual decoder implementation (see Section 3.2 for fur- n =∞. These thresholds have been designed in order to accomplish certain constraints, such as limiting the ther details). The accuracy of logistic functions after the curve-fitting process is shown in Fig. 1 for the whole set maximum instantaneous BLER (iBLER) or defining an of CQI values of LTE [2], where solid lines represent the average BLER (aBLER) target. The latter approach (based on aBLER) is the one adopted by most wireless tech- analytic BLER curves whereas simulation results of a soft output Viterbi algorithm (SOVA)-based turbo decoder nologies like LTE [1]. Therefore, our description will be focused on the aBLER scenario. [10] are marked with circles. A static selection of the values for the AMC thresh- For a certain average SNR , the average BLER under olds does not perfectly adjust the aBLER to a target AMC can be evaluated as i since link conditions are inherently variant. Thus, in order − n 1 i+1 ( { }) = AWGN (γ ) o( γ) γ aBLER , i iBLERi p , d , Table 1 Values of αi and αi of modeled iBLER curves i=0 i 0 1 CQI index α α (1) i0 i1 1 −28.08 9.71 being iBLERAWGN (γ ), the instantaneous BLER for a i 2 −20.59 11.05 given MCS i over an additive white Gaussian noise − (AWGN) channel and po(, γ) the probability density 3 15.31 12.89 function (PDF) of the instantaneous SNR conditioned to 4 −11.09 14.45 transmission. 5 −8.05 17.12 In this, work we have assumed an uncorrelated Rayleigh 6 −6.56 20.56 channel for the analysis. Thus, the PDF of the instanta- 7 −2.48 16.07 neous SNR for a certain average SNR is given by an 8 −2.39 22.83 exponential function [1] 9 −1.26 18.74 1 −γ ( γ)= / 10 −0.67 20.02 p , e .(2) 11 −0.40 18.36 Then, the instantaneous SNR conditioned to transmis- 12 −0.26 16.62 sion is given by: 13 −0.17 16.18 1 −γ/ γ> ( ) e , 0 14 −0.04 7.02 po(, γ)= A , 0 (3) 0, else 15 −0.03 8.84 Blanquez-Casado et al. EURASIP Journal on Wireless Communications and Networking (2016) 2016:20 Page 3 of 16 100 Analytic BLER for CQI 1 Analytic BLER for CQI 2 Analytic BLER for CQI 3 Analytic BLER for CQI 4 Analytic BLER for CQI 5 Analytic BLER for CQI 6 Analytic BLER for CQI 7 Analytic BLER for CQI 8 Analytic BLER for CQI 9 Analytic BLER for CQI 10 10-1 Analytic BLER for CQI 11 Analytic BLER for CQI 12 BLER Analytic BLER for CQI 13 Analytic BLER for CQI 14 Analytic BLER for CQI 15 Simulated BLER 10-2 -10-50 5 10152025303540 SNR(dB) Fig.