Population Midpoint-Based Differential Evolution for Localization in Wireless Sensor Networks

Abstract

The standard differential evolution (DE) algorithm is a prominent population-based evolutionary algorithm that has shown competitive performance in addressing highly complicated problems such as target node localization in wireless sensor networks. The DE performance, however, can be compromised when the localization problem includes additional parameters, such as the transmit power and path-loss exponent in the received signal strength, to be estimated. In this paper, we propose an enhanced DE (EDE) and its variant called vEDE for addressing the deterioration of the DE when solving the localization problem. The proposed EDE and vEDE both incorporate two processes, namely random redirection and generation of midpoint individuals, to enhance their performance. In cases of high complexity, our numerical results reveal that the EDE and vEDE improve the localization accuracy to roughly 53% and 55%, respectively, as compared to the standard DE. The results also show the superiority of the vEDE as compared to the state-of-the-art algorithms based on semi-definite programming and second-order cone programming under various settings.