Q- Learning based Localization Algorithm for Smart Agriculture in WSN-assisted IoT

• Hina Singh,
• Preeti Yadav,
• Vinay Rishiwal,
• Shashi Bhushan,
• Mohd. Shiblee,
• Mano Yadav,
• Omkar Singh,

Vol. 19, No. 12, December 31, 2025
10.3837/tiis.2025.12.016, Download Paper (Free):

• Machine Learning
• Q-learning
• Wireless Sensor Network
• Internet of Things
• Agriculture
• localization
• WIoT

Abstract

The ability of the Wireless Sensor Network (WSN)-assisted Internet of Things (IoT) to detect the exact position of nodes/devices is crucial. Smart agriculture utilizes information technology to improve crop quality and output. Using WSN, IoT, and machine learning can help create highly productive and environmentally friendly smart agriculture. Devices in WSN-assisted IoT (WIoT) networks are often spread randomly, and the cost of integrating Global Positioning System (GPS) units into them is too expensive. However, accurate location is crucial in various applications such as animal monitoring, agricultural environment data collection, underwater activity monitoring, etc. WIoT, especially those without a global positioning system, have one of the most difficult challenges in terms of localization. It does not need a specialized positioning device; a range-free localization system is a promising, cost-effective option. The proposed algorithm improves localization by using reinforcement learning methods, encouraging decision-making and adaptive sampling in a changing environment. To test the algorithm, we perform the dry run in a simulated environment, providing a real-world situation, a large area of about 100x100 m2 with populated sensor nodes in a dynamic mode. The work shows better localization accuracy performance as an error has a bracket of (0.06 m to 0.1 m), which approximates between (15 to 20) % reduction in the localization errors compared to traditional methods. The localization technique has been designed to self-tune the sampling and filtering techniques, which leads to fast convergence in the case of stable error rates that range between 15-20 iterations compared to 20-25 iterations in other cases. Furthermore, it also states that a lower energy consumption of 15-25 Joules will make energy more environmentally sustainable. Parameter scalability was found to be well-tested, with many nodes showing stability up to 500 nodes. Results confirm that the Q-learning-based localization algorithm has great potential for an efficient, scalable, and accurate solution for Intelligent agriculture applications in the WIoT environment.

Statistics

Cite this article