Q-LEARNING-BASED TRUST-AWARE CLUSTERING AND ADAPTIVE ROUTING FOR WIRELESS SENSOR NETWORKS IN THE INTERNET OF THINGS: A SIMULATION STUDY
Abstract
Wireless sensor networks (WSNs) supply much of the data used in Internet of Things (IoT) systems, but their battery-powered nodes have limited energy and are exposed to security threats. This study describes a framework in which each sensor node uses Q-learning to support cluster head (CH) selection and routing. A dedicated Checkup Node (CN) estimates trust from direct and indirect observations and moves suspicious nodes first to a gray list and then to a blacklist. Trust is a condition for CH eligibility and one input to next-hop selection. The sink forms the initial clusters with K-means and a cost function based on energy, distance, and link quality. The framework was compared with the TEERP and RLER protocols in a custom Python simulator with 100 static nodes in a 100 m × 100 m area over 1000 rounds in a single run. By round 700, the proposed framework had used 800 J of the 1000 J network energy, compared with 750 J for TEERP and 700 J for RLER. It kept more nodes alive, with 75 nodes remaining at round 1000 against 10 and 47, and its mean trust over the logged rounds was 0.90 against 0.77 and 0.79. Healthcare monitoring is discussed as a motivating application, but no healthcare scenario or medical data were simulated. No attacks, node failures, or larger networks were simulated. The results therefore show that the design is feasible under benign conditions. They do not show improved security or scalability.
Keywords
- Q-learning, Wireless sensor networks, Trust management, Cluster head selection, Internet of Things.
Views & Downloads Statistics (Last 12 Months)
Monthly performance and trends of article views and downloads
