A Maximum-Throughput Call Admission Control Policy for CDMA Beamforming Systems Wei Sheng and Steven D. Blostein Department of Electrical and Computer Engineering Queen’s University, Kingston, Ontario, Canada Emails: wsheng@ee.queensu.ca, Steven.Blostein@queensu.ca Abstract— A throughput-maximization call admission control (CAC) policy is proposed for CDMA beamforming systems in which the QoS requirements in both physical and network layers can be guaranteed. While the existing cross-layer CAC policies rely on a separate reduced-outage-probability (ROP) algorithm to guarantee the physical layer QoS requirement, which adds to system complexity and reduces spectral efficiency, the proposed CAC policy can maintain arbitrary outage probability constraints as well as all the other QoS requirements without the aid of any ROP algorithm. The optimal CAC policy, obtained by formu- lating a constrained semi-Markov decision process (SMDP), is able to optimize the overall system throughput across different layers. Numerical examples demonstrate that the proposed policy is capable of achieving a significant performance gain, in terms of lowered blocking and outage probabilities as well as increased system throughput. I. INTRODUCTION Recently, the problem of ensuring quality-of-service (QoS) requirements in both physical and network layers by designing a cross-layer CAC policy is receiving much attention. In [1]- [2], optimal semi-Markov decision process (SMDP)-based CAC polices are presented for the case of single-antenna systems, which lacks the tremendous benefits provided by multiple antennas. In this paper, we investigate the optimal CAC policy for a CDMA multiple-antenna system. In multiple antenna systems, the spatial channel response, parameterized by the angle-of-arrival (AoA) information may be employed at the receiver to suppress interference. The resulting signal-to-interference ratio (SIR) is a random process determined by the realizations of AoAs. The large fluctuations in this spatially filtered SIR can lead to a significant outage probability in the physical layer, defined as the probability that the target SIR cannot be satisfied. Outage probability constraints are discussed in [2]. However, the CAC policy in [2] considers a single antenna system, which relies on a specific large system analysis. Furthermore, existing methods for cross-layer admission control in the current literature, e.g., [1] [2], treat the SIR as quasi-static and do not work well for multiple antenna systems. Therefore, designing an optimal CAC policy for multiple antenna systems can be a very challenging problem since the outage probability must be controlled jointly with the network layer operation. In [3] [4], suboptimal CAC policies are derived for CDMA beamforming systems, in which a separate reduced-outage- probability (ROP) algorithm is required to mitigate the outage probability. Several efficient ROP algorithms are proposed in [3], which can reduce the outage probability to a tolerably small level. These ROP algorithms, however, either introduce cost in system resources, such as reduced spectral efficiency and increased computation complexity, or degrade the network layer performance [3]. Furthermore, in [1]- [4], only network layer optimization is considered, which is inferior to the CAC policy jointly optimized across physical and network layers. This motivates our research on a throughput-maximization CAC policy for multiple antenna systems without the aid of a ROP algorithm. An exact outage probability is derived in the presence of both voice activity and multiple antennas. Based on this outage probability, the optimal CAC problem is obtained by formulating a constrained semi-Markov decision process, which can guarantee arbitrary outage probability constraints without the aid of a ROP algorithm. The proposed policy optimizes the overall system throughput across physical and network layers. To the best of our knowledge, the CAC design which maximizes the overall system throughput across different layers has not been addressed in the literature. To highlight the maximum-throughput CAC design, error- control schemes such as automatic retransmission request (ARQ) are ignored in this paper. However, in our companion paper [5], an optimal admission control (AC) policy as well as a low-complexity suboptimal version are developed that incorporate a truncated ARQ scheme [5]. In summary, this paper differs from [5] in the following ways: a) In this paper we study a connection-oriented network in which voice activity factor is employed to increase the user capacity, while in [5], a connectionless communication is assumed in which no activity factor is considered; b) In this paper we focus on the optimal CAC design by incorporating the outage probability as well as all the other QoS constraints into the Markov decision process, while in [5] we mainly emphasize the formulation of the constrained admission control problem by considering the impact of ARQ. The rest of this paper is organized as follows. The signal model and problem formulation are presented in Section II. Section III investigates the physical layer performance and provides an analytical expression for outage probability. Optimal CAC policies for multiple-class systems are proposed in Section IV. Numerical results are presented in Section V. 1525-3511/08/$25.00 ©2008 IEEE This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings. 2986 Authorized licensed use limited to: Queens University. Downloaded on May 2, 2009 at 19:47 from IEEE Xplore. Restrictions apply.