Fitness approximation with RF algorithm dedicated to WSN node deployment for a soil monitoring application. (2nd June 2023)

• Record Type: Journal Article
• Authors: Taleb, Soumaya Ferhat; Benalia, Nour El-Houda; Sadoun, Rabah
• Is Part Of: International journal of ad hoc and ubiquitous computing. Volume 43:Number 2(2023)
• Page Start: 72
• Page End: 86
• Publication Date: 2023-06-02
• Subjects: precise agriculture -- wireless sensor network -- WSN -- node deployment -- fitness approximation -- random forest -- strength Pareto evolutionary algorithm II -- SPEA II; Ubiquitous computing -- Periodicals; Embedded computer systems -- Periodicals; Electronic data processing -- Distributed processing -- Periodicals; Wireless communication systems -- Periodicals; Computer architecture -- Periodicals; 004.2
• Languages: English
• ISSNs: 1743-8225

Abstract

In order to solve the wireless sensor network (WSN) node deployment optimisation for an agricultural application, a hybridised strength Pareto evolutionary algorithm II with the random forest regressor (RF-SPEA II) was used. The SPEA II intended to optimise the deployment according to the classical constraints of coverage, over-coverage, connectivity and node number, in addition to the nodes separating distance constraint, which affects the predicted physical parameters models quality. Furthermore, the RF regressor was applied as a fitness approximation surrogate model, with the use of evolutionary control rate to avoid convergence to false optimums. Moreover, the application of RF features selection skill that helped to only keep important characteristics and gain more time. Consequently, this hybridisation allowed finding results that exceeded the unaltered SPEA II in terms of solutions qualities and computational time. For example, for an agricultural plot of 400 m 2 of surface, the RF-SPEA II hybridised algorithm gave better constraint rates and was 5.82 times faster than the unaltered SPEA II.