New resource management strategy for wireless cellular networks

Abstract

In this paper, the fuzzy Hopfield neural network (FHNN) is proposed for channel assignment in wireless cellular system. Each channel is regarded as a data sample and every cell is taken as a cluster. Channels are adequately distributed to the dedicated cells while satisfying the interference constraints such as co-site constraints, adjacent channel constraints, and co-channel constraints. The goal is to avoid the interference and serve the expected traffic, which is to minimize used spectrum. Moreover, interference prediction is a delicate task; it depends on the details of the traffic assumptions. The FHNN guarantees that the neural network will skip local minima, and in all cases will converge to the optimum arrangement of the channels. Simulation results show that the FHNN can provide an alternative approach of solving this class of channel assignment problems.

Introduction

Due to the usable range of the frequency spectrum is very limited, channel assignment problems (CAP) become very important to improve traffic performance in mobile communication systems. The operator has to assign channels to base stations, in a way that the quality of service (QoS) is guaranteed. It is an NP-complete optimization problem for designing of cellular radio system. The objective of selecting a channel assignment scheme [1] is to achieve a high degree of utilization. Channels are assigned to different neighboring cells to reduce interference and increase overall system capacity [2], [19].

Many researchers have investigated neural networks to solve the CAP [3], [4], [5], [6], [7], [8]. In 1991, Kunz [3] first applied the Hopfield neural network model to solve the CAP in the cellular radio network. The Kunz’s neural-network model required a large number of iterations in order to reach the final solution. Funabiki and Takefuji [4] proposed a neural network model composed of the hysteresis McCulloch–Pitts neurons, and four heuristics were used to improve the convergence rate of channel assignment. They also tried to fix the channel assignment in one or more certain cells in order to accelerate the convergence time. Kim et al. [5] proposed a modified discrete Hopfield neural network algorithm to escape local minima problem. Genetic algorithms have also been used to solve channel assignment problems. Kaminsky [7] used genetic algorithm to solve channel assignment without scaling up the problem. Genetic algorithm performed well for small problems, but it consumed much long time and didn’t ensure to get a global optimum solution. Smith and Palaniswami [6] proposed two different neural networks, an improved Hopfield neural network and self-organizing neural network, for solving the CAP. He et al. [9] proposed a combination of multistage self-organizing channel assignment algorithm and transiently chaotic neural network to solve the cellular channel assignment. Unfortunately, in the case of the CAP, He et al. method takes much calculation time to choose the optimum parameters. An inherent disadvantage of the approach is that it easily converges to the local optima and, hence, to get suboptimal solutions only. To overcome this disadvantage, a simulated annealing approach was suggested by Duque-Antn et al. [10]. Although the approach proposed by Duque-Antn et al. is guaranteed to achieve the global optimum asymptotically, the convergence speed is rather slow, and a carefully designed cooling schedule is required.

Most schemes employed to solve CAP problems are concentrated on the applications of linear programming, graph coloring algorithm, genetic algorithm [7], [8], annealing [9], neural networks [3], [4], [5], [6], and fuzzy logic [10], [11], [12]. No attempt has been made to solve CAP by applying a fuzzy clustering technique. The problem of fuzzy clustering is that it is needed to find a fuzzy c-partition and the corresponding cluster centers. In this study, fuzzy Hopfield neural network (FHNN) clustering technique is applied to solve the CAP. The proposed algorithm integrates fuzzy c-means clustering strategies into learning procedures [14], [15]. The FHNN algorithm is a numerical procedure for finding membership grade, which minimizes the energy function. Therefore, the CAP is then considered as the problem of minimizing an energy function. The FHNN algorithm is firstly used to obtain the net values, then the fuzzy states updating procedure is applied to obtain the solution. Channels are adequately distributed to the cells while satisfying the interference constraints. The FHNN guarantees that the neural network will skip local minima and, in all cases, be converged to the optimum arrangement of the channels. This approach is based on an unsupervised two dimensional fuzzy Hopfield neural networks. Lyapunov energy function is used as the objective function and the CAP problem is become to solve the optimum problem depending on the Lyapunov energy function. The proposed approach will increase the convergence speed that results to reduce the number of iterations.

The rest of this paper is organized as follows. In Section 2, the channel assignment problem is described. In Section 3, the mathematical representation of the fuzzy Hopfield neural network (FHNN) is firstly described, and then how to apply the algorithm to solve channel assignment problem is presented. Section 4 shows the simulation results. Finally, conclusions are presented in Section 5.