Hybrid Memetic Algorithm for the Node Location Problem in Local Positioning Systems

Abstract

Local Positioning Systems (LPS) have shown excellent performance for applications that demand high accuracy. They rely on ad-hoc node deployments which fit the environment characteristics in order to reduce the system uncertainties. The obtainment of competitive results through these systems requires the solution of the Node Location Problem (finding the optimal cartesian coordinates of the architecture sensors). This problem has been assigned as NP-Hard, therefore a heuristic solution is recommended for addressing this complex problem. Genetic Algorithms (GA) have shown an excellent trade-off between diversification and intensification in the literature. However, in Non-Line-of-Sight (NLOS) environments in which there is not continuity in the fitness function evaluation of a particular node distribution among contiguous solutions, challenges arise for the GA during the exploration of new potential regions of the space of solutions. Consequently, in this paper, we first propose a Hybrid GA with a combination of the GA operators in the evolutionary process for the Node Location Problem. Later, we introduce a Memetic Algorithm (MA) with a Local Search (LS) strategy for exploring the most different individuals of the population in search of improving the previous results. Finally, we combine the Hybrid Genetic Algorithm (HGA) and Memetic Algorithm (MA), designing an enhanced novel methodology for solving the Node Location Problem, a Hybrid Memetic Algorithm (HMA). Results show that the HMA proposed in this article outperforms all of the individual configurations presented and attains an improvement of 14.2% in accuracy for the Node Location Problem solution in the scenario of simulations with regards to the previous GA optimizations of the literature.

Memetic Chains for Improving the Local Wireless Sensor Networks Localization in Urban Scenarios

“…The MAs have been applied in many different problems in the literature [ 66 ]. In the optimization of the NLP in WSN, MAs have been previously considered for the coverage problem [ 67 , 68 ] and we later extend this procedure to include the positioning accuracy in the localization NLP presented in Section 2 considering both LOS and NLOS links in the signal paths [ 44 ].…”

“…Thus, we introduce some changes in the base algorithm (i.e., MA-SW-Chains) for achieving practical results in this problem through a novel algorithm called MA-Variable Neighborhood Descent-Chains (MA-VND-Chains): The introduction of the VND LS in the MA-VND-Chains instead of the SW LS of the MA-SW-Chains for exploring a discrete neighborhood around the individual of the GA selected for the LS. The utilization of different knowledge for the selection of the individuals of the LS in the MA-VND-Chains through a different definition of the LS Intensity based not only on the improvement of past LS iterations (i.e., MA-SW-Chains) but also in the information of the size of the neighborhood analyzed around each of the nodes of the NLP and diversity criteria through the number of LS iterations applied previously to the individual selected which is critical to address discontinuous optimization as we previously showed in [ 44 ]. MA-SW-Chains presents a steady-state GA which is not optimal for the NLP since an extensive exploration of the space of solutions is required for achieving practical results.…”

“…The utilization of different knowledge for the selection of the individuals of the LS in the MA-VND-Chains through a different definition of the LS Intensity based not only on the improvement of past LS iterations (i.e., MA-SW-Chains) but also in the information of the size of the neighborhood analyzed around each of the nodes of the NLP and diversity criteria through the number of LS iterations applied previously to the individual selected which is critical to address discontinuous optimization as we previously showed in [ 44 ].…”

“…The definition of the most diverse individuals is achieved throughout the measurement of dissimilarity among solutions. Therefore, a comparative between every sensor location among individuals is performed, as presented in [ 44 ].…”

“…Therefore, heuristic methods have been applied in the literature for solving this complex problem in the field of the Wireless Sensor Networks (WSN). Genetic algorithms (GA) [ 34 , 35 ], the firefly algorithm [ 36 ], simulated annealing [ 37 ], the bacterial foraging algorithm [ 38 ], the grey wolf optimizer [ 39 ], the bat algorithm [ 40 ], tabu search [ 41 ], the elephant herding optimization [ 42 ], differential evolution [ 43 ], or memetic algorithms (MA) [ 44 ] have been applied for achieving practical results for the NLP.…”