A Localization Algorithm Joining DV-Hop, LSSVM, and Expected Distance Estimation in IoT of Agriculture in Mountainous and Hilly Areas

The problem of wireless sensor networks (WSNs) node localization is one of the hottest research topics in the application of internet of things (IoT). DV-Hop has recently significantly pushed the state of the art in node localization algorithms. However, DV-Hop and its variants cannot achieve excellent performance, when estimating distance independently in the application of WSNs deployed in mountainous and hilly areas. Therefore, this article proposes a novel node localization approach that combines DV-Hop, least squares support vector machine (LSSVM), and expected distance estimation in order to effectively calculate the estimated distance of nodes in IoT of agriculture. First, the estimated expected distance is proposed based on the different number of hops from the unknown node to the beacon node. Second, a regression model based on LSSVM is applied to predict the distance between the unknown node and the beacon node. Additionally, three objective functions for joint estimation of localization information are constructed. Finally, multiobjective golden eagle optimization (MOGEO) is utilized to solve the coordinates of the unknown node. The experimental results indicate that the average positioning error (APE) of the proposed approach is promising, and reduces node localization error by 39.6% on average with APE compared to the original distance-vector hop (DV-Hop) and its variants in randomly distributed networks.