Abstract:
Recently, Wireless Mesh Network (WMN) has gained important roles in current communication technologies. It has been used in several applications and most of them are critical applications such as surveillance, transportation systems and rescue systems. Hence, the WMN attracts a lot of attention from many researchers. WMN consists mainly of mesh clients MCs and mesh routers MRs, some of the latter are supplied by additional functions to serve as Internet gateways (IGs). Thus, most of the network traffic is acting toward IGs. Therefore, the network performance largely depends on the MRs’ placement, especially the IGs. There are many research efforts on solving the gateway placement problem (GPP) and it has been proven to be NP-Complete by many researchers. Thus, finding the optimal solution is difficult. Therefore, finding near optimal solution is crucial to improve the network performance. This research proposes a novel approach to solve this problem using Genetic Algorithm (GA) and Simulated Annealing (SA) to achieve a near optimal solution guided by a mathematical model, considering the number of IGs and the number of hops that a packet traverses between the IG and the source / destination MR (MR-IG). The main objective of the proposed approach is to minimize the variation of MR-IG-hop counts (VAR-MR-IG-Hop) among MRs to insure that the IGs are placed in the appropriate positions. Finally, the proposed set of algorithms is evaluated using many generated instances using different parameters (population size, tournament size, crossover type, mutation type) for GA and many parameters for the SA such as the internal temperature, the final temperature and the parameters that have been used to change the internal temperature and in the transition function. Furthermore, a comparison between the two algorithms have been done. The experimental results for GA have shown high convergence rate. Moreover, the algorithm has considerable scalability and robustness to solve the GPP in large and small networks. In addition, SA has shown high convergence rate and fast execution time in comparison with the GA. However, GA has better performance in the small-size network with high scalability opportunities while SA is faster than GA in the large-size network but it has limited chances for further optimization.
