Coverage Hole Detection using Deep Learning Models

Abstract

Wireless Sensor Networks (WSNs) function as essential components for contemporary systems because they serve multiple applications including environmental tracking and disaster control systems and modern smart urban planning and military field surveillance. WSNs face an essential challenge because sensor nodes are deployed randomly while environmental factors and battery drain and physical damage lead to node failures. Coverage holes which are areas without sensor monitoring create severe performance problems that affect data collection and routing efficiency and energy usage. Modern research focuses on three approaches for hole detection and repair in WSNs which involve computational geometry as well as statistical paradigms and topological analysis methods. The established techniques of Voronoi tessellation and Delaunay triangulation efficiently work on structured conditions yet fail to handle dynamic or irregular terrain situations.

Deep learning solutions provide an effective answer to these difficulties through spatial-temporal hole identification capabilities. Both Convolutional Neural Networks (CNNs) extract spatial features present at node deployment maps and Transformers interpret temporal patterns that include node failures over time. The use of reinforcement learning approach enables optimal mobile node deployment through efficient movement to fix coverage deficits in real-time. This proposed algorithm combines the mentioned techniques to deliver advanced performance in detection accuracy with energy efficiency and better scalability for multiple environments. Results from simulations show that the proposed method improves current coverage hole solutions effectively by increasing coverage to access 98.6% while demonstrating consistent performance in resolving this widely known WSN problem.