4576 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 58, NO. 8, OCTOBER 2009 Unified Approach to Joint Power Allocation and Base Assignment in Nonorthogonal Networks Lev Smolyar, Itsik Bergel, and Hagit Messer Abstract—In this paper, we consider the joint power allocation and base station (BS) assignment in nonorthogonal downlink transmission code-division multiple-access (CDMA) communica- tion systems. In addition to the popular sum-power criterion, we introduce three novel criteria that better take into account the network requirements and particularly the separate power limit of each of the BSs. In our model, we allow each user to be connected to more than one BS and show that such a scheme can improve the system capacity (while currently this mechanism is only used to support soft handoff). Nevertheless, we show that in any optimal solution, the actual number of connections used by each user is small. We propose a unified approach to optimization, which allows the use of linear programming (LP)-based iterative algorithms to solve all of the presented optimization problems. The proposed scheme is verified by simulations for several scenarios. The simulation results show that by taking into account the power constraints of each of the BSs, the system capacity can be increased by up to 33%. The results can also approximate the performance of orthogonal CDMA networks with precision that increases with system bandwidth. Index Terms—Base assignment, code-division multiaccess (CDMA), power allocation, power control. I. I NTRODUCTION N EW generations of wireless networks place high demands on multimedia applications, such as video streaming and Internet, together with classical applications, such as voice communication. These services impose high requirements, par- ticularly on the downlink throughput. As radio resources are often scarce, careful and efficient allocation of the limited resources is vital. In many multiple-access systems (for exam- ple, code-division multiple access (CDMA) [1]–[3], frequency- division multiple access, time-division multiple access [4], and impulse radio (IR) [5]), the same spectrum is shared by multiple cells. Thus, the power transmitted to each user is interference to other users. This interference is one of the most limiting factors on system capacity. The level of interference between the users depends on the algorithms used to assign users to base stations (BSs) and for power allocation (also termed power control). Manuscript received October 21, 2007; revised October 26, 2008 and January 19, 2009. First published April 7, 2009; current version published October 2, 2009. The review of this paper was coordinated by Prof. B. L. Mark. L. Smolyar and H. Messer are with the School of Electrical Engineering, Tel-Aviv University, Tel-Aviv 69978, Israel (e-mail: smolyar@eng.tau.ac.il; messer@eng.tau.ac.il). I. Bergel is with the School of Electrical Engineering, Tel-Aviv University, Tel-Aviv 69978, Israel, and also with the School of Engineering, Bar-Ilan University, Ramat-Gan 52900, Israel (e-mail: ibergel@eng.tau.ac.il). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TVT.2009.2020330 The work presented herein may be applicable to most multiple-access schemes [6]. However, for convenience, it is presented in the framework of CDMA, where the concept of nonorthogonal transmission is more relevant [7]. Note that the analysis of IR systems is closely related to that of CDMA sys- tems [8]–[10] and, hence, the work herein is directly applicable for IR systems as well. Nonorthogonal CDMA downlink transmission is less ef- ficient than orthogonal transmission because it creates more interference. However, nonorthogonal systems are more readily analyzed, and therefore, they were widely used to evaluate lower bounds on orthogonal system performance (e.g., [7, Sec. 6.7]). Moreover, in wideband systems, signal orthogo- nality is corrupted by multipath scattering, and therefore, the performance of an orthogonal system is very close to that of a nonorthogonal system. Due to the relative simplicity of the uplink, most of the research had concentrated on uplink scenarios (e.g., [11] and [12]), and a comprehensive summary can be found in [13]. These works assumed nonorthogonality between different users of the same cell. They cannot directly be applied to the down- link, where the constraints are on the sum of power allocated to all users in a given cell (in contrast with the uplink, where the constraint is on the power allocated to each of the users). The research on the downlink channel in multicell systems mostly assumed that the user allocation was predetermined [14]–[18]. Berggren et al. [19] analyzed an orthogonal system, whereas Kim [20] bounded the performance of an orthogonal system by the performance of a nonorthogonal system. Other works suggested suboptimal user assignment approaches. For exam- ple, Lee et al. [21] considered the maximization of the expected throughput by joint resource allocation and BS assignment in an orthogonal system. The BSs independently allocate the power and the rate based on the user-to-base assignments from the previous stage. At the next stage, if the load is not balanced, then they reassign some users from heavily loaded cells to lightly loaded cells. Optimal joint user assignment and power allocation for the downlink was only presented for the sum-power criterion, i.e., for the minimization of the sum of all BS transmitted power. Rashid-Farrokhi et al. [22] utilized the uplink–downlink duality principle, stating that, for a given user-to-base assignment, the sum of the uplink powers is equal to the sum of the downlink powers (up to a constant). Using this duality principle [22], the users in the downlink were assigned according to the uplink assignment previously developed. After the users were assigned, the power that achieved the required performance for each user was determined. 0018-9545/$26.00 © 2009 IEEE