Connection Admission Control Reward Optimization for Different Priority Classes in Homogeneous Wireless Network Amr Adel Nasr, Hesham Mohamed El-Badawy Network Planning Department National Telecommunication Institute Cairo, Egypt E-mail: amr.adelnasr@gmail.com, hesham@nti.sci.eg Hadia M. El-Hennawy Electronics and Communication Department Faculty of Engineering, Ain Shams University Cairo, Egypt E-mail: helhennawy@ieee.org Abstract— Nowadays, Connection Admission Control (CAC) reward optimization for a service with different priority classes in a homogeneous wireless network has become one of the most important issues. In this paper, a novel expansion for the Guard Channel Based Incremental and Dynamic Optimization algorithm is presented. GUIDO will be used to analyze a reward scheme that correlates service demand with weights on pricing, after which optimal weights are deduced for maximum system reward, taking into consideration QoS for the different priority classes. An analytical model is presented based upon Lagrangian Optimization criterion and is used to verify the optimal weights. The paper thus, suggests the optimal weights on pricing for maximum system reward and offers an analytical derivation for them. Keywords- GUIDO; CAC; MDP; Optimal Weights; Lagrange Optimization I. INTRODUCTION Connection Admission Control (CAC) is a technique to provide QoS in a network by restricting the access to network resources. For given free radio resources, CAC will provide the most suitable call acceptance scheme in order to guarantee the committed QoS for the already accepted calls. There is a tradeoff between the QoS level perceived by the user (in terms of the call dropping probability) and the utilization of scarce wireless resources. In fact, CAC can be described as an optimization problem [2]. In a homogeneous wireless network, two types of connection requests exist. Namely they are new calls and handoff calls. Since dropping an ongoing call might disappoint customers, handoff calls must be assigned higher priority. To prioritize handoff calls over new calls, some channels (referred to as the guard channels) are reserved for handoff calls [3]. Specifically, if the total number of available channels is C and the number of guard channels is C-K, a new call is accepted if the total number of channels used by ongoing calls is less than the threshold K, while a handoff call is always accepted if there is an available channel [4]. In the current paper, the Guard- Channel-based Incremental and Dynamic Optimization (GUIDO), which was introduced in [1], will be used. GUIDO is a threshold-based CAC scheme. This threshold is dynamic according to traffic conditions. In [1], GUIDO was deployed in order to maximize the system reward for given system resources. However, this maximization process aimed at allocating system resources given that fixed weights (w n weight on pricing for new calls & w h weight on pricing for handoff calls) have been assigned by the service provider. So, it is required to address the issue of determining optimal weights [5]. In consistence with the pricing scheme suggested by the empirical study [6] that relates the arrival rate and its weights, the current paper will utilize the GUIDO algorithm to derive optimal weights for a single service with multiple priority classes in a mobile wireless network for reward optimization with quality of service (QoS) guarantees. This is considered as an enhancement and more realistic model for GUIDO algorithm. Now to solve our CAC problem, we need some sort of a stochastic sequential decision model. One well known stochastic sequential decision model is the Markov Decision Process (MDP) [7]. In MDP, the set of available actions, the rewards, and the transition probabilities depends only on the current state and action and not on states occupied and actions chosen in the past. The model is sufficiently broad to allow modeling most realistic sequential decision-making problems. The MDP can be used to determine the optimal call-admission policy for priority treatment as in [8], [9], [10], [11]. In most cases, where the traffic condition does not change rapidly, the MDP-based call-admission policy provides a better tradeoff between optimality and complexity. MDP models are widely used in various research areas and applications [12], [13], [14]. The goal of this paper is to extend the work done in GUIDO [1], taking into considerations the proposed pricing scheme to derive the optimal weights for new and handoff calls. In addition, Lagrange Optimization Method will be used similar to the method adopted in [15] in order to verify the optimal weights derived. The rest of the paper is organized as follows. In Section II, the system model will be presented including a validation of the work previously done in [1] and the novel expansion to [1] using the suggested pricing scheme. Section III presents the numerical results with physical interpretation given. Finally, Section IV offers a conclusion, a summary of the work done and outlines our future research directions.