Flexible Adaptive-Modulation-and-Coding Tables for a Wireless Network Edward W. Jang * , Chan-Soo Hwang † , and John M. Cioffi * * STAR Laboratory, Stanford University, Stanford, CA 94305, U.S.A. Email: {ej1130, cioffi}@stanford.edu † Communication & Network Lab., Samsung Advanced Institute of Technology, Yongin, Korea Email: cshwang@ieee.org Abstract— This paper proposes to use flexible adaptive- modulation-and-coding (AMC) tables in a wireless network. To support the flexibility of AMC tables, a low-complexity AMC table optimization algorithm is also proposed. The proposed algorithm iteratively optimizes the switching levels of an AMC table according to channel environments, and has fast conver- gence speed. Computer simulation results show that the proposed algorithm optimizes AMC tables according to different cell characteristics, and that using flexible AMC tables for a wireless network achieves near-optimal average spectral efficiency even with a small feedback load. I. I NTRODUCTION Many recent cellular systems, such as IEEE 802.16e stan- dard [1], employ an adaptive-modulation-and-coding (AMC) scheme to adjust the transmission rate according to the chan- nel state information (CSI), thereby achieving high spectral efficiency while guaranteeing sufficient reliability [2]. Specif- ically, an AMC scheme assigns the CSI intervals, bounded by switching levels, to combinations of a modulation size and a code rate based on an AMC table. When a user’s CSI falls in an interval, the user feeds the interval’s index back to the base station (BS), and the BS transmits a packet to the user with the corresponding transmission rate. With more switching levels in the AMC table, the spectral efficiency increases because round-off errors decrease; however, the feedback load also increases with the number of switching levels. This feedback overhead is especially problematic for an orthogonal-frequency-division multiple-access (OFDMA) system, where each user needs to feed AMC table indexes for multiple subchannels. Therefore, a judicious choice of the AMC table’s switching levels is crucial to increase the spectral efficiency while minimizing the feedback load. The optimal set of switching levels of an AMC table crucially depends on many channel characteristics, such as a users’ geographical distribution, a coverage radius, path loss, shadowing, and transmit power. Since a wireless network often comprises channels with different characteristics, e.g., a cellular system has urban microcells as well as suburban macrocells, employing flexible AMC tables optimized for dif- ferent channel characteristics significantly improves the overall spectral efficiency of a wireless network compared to the case of employing a fixed AMC table across channels with different characteristics. However, using flexible AMC tables requires an efficient algorithm that optimizes the switching levels for arbitrary channel state distributions. Currently applicable algorithms require a large computa- tional complexity to optimize an AMC-table, i.e., to find the switching levels that maximize the average spectral efficiency. This optimization problem was first considered by Holm et al., who found a recursive solution for a Rayleigh fading channel in [3]. However, the AMC-table optimization for a general channel state distribution has remained an open problem. Other algorithms can be applied by relating this AMC-table optimization problem to the random variable quantization problem. However, the widely-known Lloyd algorithm [4] or the algorithms using sufficient conditions for the optimality [5] are not straightforwardly applicable because the distor- tion measure for this optimization problem is asymmetric. The dynamic-programming-based quantization algorithm [6], which includes asymmetric distortion measure cases, is appli- cable for this optimization problem; however, its complexity is prohibitively high due to the lack of a boundary condition for switching levels. To solve this problem, this paper proposes an iterative flexible-AMC-table optimization (IFAO) algorithm that itera- tively finds one switching level of an AMC table per step by using the necessary conditions for the optimality. This iterative approach significantly reduces the computational complexity compared to other algorithms that simultaneously optimize all switching levels [3], [6]. This is because using only the necessary conditions reduces the computational complexity that would have incurred if both the necessary and sufficient conditions were used to solve the problem. Simulation results show that the proposed IFAO algorithm optimizes an AMC table that achieves the average spectral efficiency, very close to the unlimited feedback case, with only a small feedback load. The simulation results also show that using flexible AMC tables, instead of using a fixed AMC table, significantly improves the average spectral efficiency of a wireless network. II. SYSTEM MODEL AND PROBLEM STATEMENT The system model considers a downlink channel, consisting of a BS with one transmit antenna and multiple users, each with one receive antenna. Although the channel between the BS and a user is frequency-selective, OFDMA divides the channel into multiple frequency-flat subchannels that are This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 proceedings. 978-1-4244-2075-9/08/$25.00 ©2008 IEEE