Scheduling in OFDMA Systems With Outdated Channel Knowledge Hassan Ayoub and Mohamad Assaad Ecole Superieure d’Electricite (SUPELEC) 3 rue Joliot-Curie 91192 GIF sur YVETTE - France {Hassan.Ayoub, Mohamad.Assaad}@supelec.fr Abstract— In wireless systems, most of nowadays scheduler assumes perfect knowledge of the channel quality indicator (CQI) at the transmitter. Unfortunately, this is never the case since the receiver (e.g. mobile users) estimates the CQI at time t - τ and feed back the estimate to the transmitter (e.g. base station). The transmitter uses then this outdated CQI at time t to allocate the radio resources to the users. This paper analyzes this issue at various mobile velocities and proposes improvement of scheduling that reduces the impact of outdated CQI. Using a channel prediction model, we determine the probability distribution function (pdf) of CQI conditioned on the outdated CQIs. We use then this (pdf) in the evaluation of the Block Error Rate (BLER) of the users’ transmission and we exploit the result in the development of a novel scheduling rule. Simulation results show improvement of the system performance without any fairness loss. 1 I. I NTRODUCTION OFDMA is a very promising radio access technology that has been adopted for both uplink and downlink air interfaces of WiMAX fixed and mobile standards, namely IEEE802.16d and IEEE802.16e respectively [1] [2], and more recently for the downlink air interface of the Third Generation Partnership Project (3GPP) currently normalizing the Long Term Evolu- tion (LTE) of the third generation (3G) cellular system [3]. In OFDMA systems, the users are multiplexed in frequency where each user’s data is transmitted on a subset of the sub-carriers of an orthogonal frequency division multiplexing (OFDM) symbol. In order to achieve the challenging spectral efficiency and user throughput targets, fast link adaptation (Adaptive Modulation and Coding) is tightly coupled to Scheduling (resource allocation scheme). The resource allo- cation exploits available channel state information (CSI) at the transmitter side for accomplishing an adaptive modulation and shares the resources (sub-carriers, slots) among users. The problem of resource allocation in OFDMA systems has been an area of active research and several scheduling algorithms have been proposed. These works considered var- ious formulations dealing with various fairness criteria (e.g. max-min fairness, proportional data rates and utility-based fairness). In [5], the authors determine the number of sub- carrier to allocate to each user using an algorithm that balances the trade off between the channel state information and the packet delay information. The sub-carrier assignment problem 1 This work was conducted in the framework of French national project APOGEE. is solved by an algorithm that monitors the violations of the maximal delay in all queues and by dividing the violations occurrences among all users. The author in [6] proposes a novel scheduling algorithm for video streaming by introducing various QoS and CQI metrics. In [7], a Proportional Fair PF algorithm is proposed in the context of TDMA systems. In [8], a multi-carrier PF (called MPF) scheduling is proposed for OFDMA systems (LTE and WiMAX). It consists in allocating each sub-carrier to the user having the highest instantaneous bit rate relative to its average achieved rate. In [9], a scheduling algorithm based on an exponential rule is proposed. In [10], a weight factor is introduced in the PF scheduling rule in order to handle multi service traffic. The obtained scheduler is called Modified Largest Weighted Delay First (MLWDF). Note that although exponential rule and MLWDF are proposed for TDMA systems, extension of these algorithms to the context of OFDMA systems is straightforward. In [11], a gradient utility based scheduling for multi-service traffic in OFDMA systems is proposed. A common underlying assumption among these works is that the channel quality indicators (CQI)s of the users are known perfectly. This assumption is quite unrealistic due to channel feedback delay. In fact, the users estimate the CQI at time t - τ and feed back these CQI estimates to the base station. The scheduler uses then these outdated CQIs at time t to make the allocation. In this paper, we analyze the impact of outdated CQI on the performance of MPF scheduling in the context of downlink OFDMA systems. This algorithm is widely used and referenced in the state of the art. We propose a modification of this algorithm in order to deal with feedback delay. We proceed first by determining the probability distribution function (pdf) of CQI with respect to the outdated available CQIs and modify then the scheduling rule using this pdf. The improvement method proposed in this paper can be applied to any other scheduling rule. The remainder of this paper is organized as follows: section II provides the problem statement and system description. In section III, the analysis and proposed scheduling improvement are provided. Simulation and results are provided in sections IV and V. Section VI concludes the paper. II. PROBLEM STATEMENT AND SYSTEM DESCRIPTION The OFDMA access technology considered in this paper re- lates to the downlink of 3GPP/LTE like system. In 3GPP/LTE