Node Distribution Optimization in Positioning Sensor Networks through Memetic Algorithms in Urban Scenarios

Abstract

Local Positioning Systems (LPS) are dependent on environmental characteristics, requiring an ad-hoc node deployment for each particular scenario of application for achieving practical results. Nonetheless, this Node Location Problem (NLP) has been assigned as NP-Hard, thus the application of heuristic algorithms is recommended for obtaining adequate solutions. Genetic Algorithms (GA) are widespread throughout the literature for solving NP-Hard combinatorial problems such as the NLP. However, GAs require the adjustment of a considerable amount of hyperparameters and can be easily compromised by premature convergence into local maximums. Therefore, in this paper, an approach based on local search methodologies, along with the GA optimization, is proposed. For this task, we apply a Memetic Algorithm based on a pseudo fitness function for reducing the problem complexity which analyses the neighboring solutions and introduces information into the optimization process. The exhaustive examination in a reduced space of solutions of this combination is idoneous for particularly adverse scenarios, thus improving the base optimization of the GA. We also perform a comparison of our method with different literature optimizations. Finally, we study the performance of the Memetic Algorithm (MA) proposed for different application scenarios, proving the effectiveness of our approach for irregular outdoor and urban context in which Non-Line-of-Sight (NLOS) conditions are considered.