Fast Wide Area Network Design Optimisation Using Differential Evolution

Abstract

The Radio Network Design (RND) problem for wide area networks consists of determining the optimal locations for base station transmitters in order to get a maximum coverage area with a minimum number of transmitters. Because of the grand amount of possible solutions, this problem is most suitable to be tackled with bio-inspired techniques instead of classical approaches. Our recent research work exploited a differential evolution based algorithm to deal with this problem. This paper describes an enhanced implementation of our Differential Evolutionbased algorithm that is able to tackle the RND problem in a most effective and faster way that the one observed in our previous work. This approach includes the comparison of different operators and the usage of an alternate initialisation model.

On the efficiency of the binary flower pollination algorithm: Application on the antenna positioning problem

“…With respect to these findings, all the works achieved after treating the antenna positioning problem used the value ˛ = 2 [13,18,[26][27][28]30]. Recently, many other works treating the antenna positioning problem used the value ˛ = 2 as a default value [45,61,62,[88][89][90][91]103].…”

“…3 [13,27,28,30,61,62,68,89] for synthetic data. The parameters of realistic data are extracted from [13,18,20,30,45,88,90,103]; and for the randomly-generated data, the parameters are extracted from [105].…”

“…Synthetic instances are provided by the University of Laguna, Spain. They were used in several works [13,28,30,61,62,89] and can be found in [4,5]. Randomly-generated instances are used too, they are provided by the University of Constantine II, Algeria.…”

“…A faster Differential Evolution algorithm with more suitable operators for the APP was proposed in [89], then Simulated Annealing (SA) and the Population-Based Incremental Learning (PBIL), the canonical Differential Evolution (DE) and an evolutionary-based algorithm (CHC) were used in [61] to solve it. In [28], the authors used a Simulated Annealing (SA), a Steady State and Generational Genetic Algorithms (SSGA, GenGA) and an evolutionary-based algorithm (CHC) to tackle the APP.…”

“…The proposed variants NI-BFPA, N-BFPA, AM-BFPA and SF-BFPA, are compared to two of the top-ranked state-of-the-art metaheuristics used to solve the APP, the Population-Based Incremental Learning (PBIL) and the Differential Evolution algorithm. The PBIL was proposed in [82] and used to solve the APP in [88,90], while the DE was first proposed in [78] and used to solve the APP in [61,62,[88][89][90][91]. Several empirical tuning experiments allowed us to find out that for values of ∈ [1, 2] the mean of the fitness value is altered only by (1/100) or (1/1000) of the original fitness value.…”