LLM-enabled adaptive scheduling in IoT sensing for optimized network performance.

Area of Science:

• Computer Science
• Artificial Intelligence
• Internet of Things

Background:

• The Internet of Things (IoT) environment is becoming more intelligent with edge devices and AI.
• Challenges in remote sensing include data incompleteness, communication delays, and dynamic topology.
• Existing systems struggle with efficient data handling and resource optimization.

Purpose of the Study:

• To propose a novel scheme, LLM-Enabled Adaptive Scheduling in IoT Sensing for Optimized Network Performance (LLM-AS).
• To leverage Large Language Models (LLMs) for adaptive scheduling in IoT sensing.
• To enhance decision-making and optimize network resources in IoT systems.

Main Methods:

• LLM-AS is trained on diverse datasets including packet loss, time fluctuations, event triggers, network failures, and congestion signals.
• The scheme is deployed in a dynamic remote monitoring system for real-time learning and feedback utilization.
• LLM-AS adjusts system sensing to prevent redundant data transmission and improve resource allocation.

Main Results:

• LLM-AS demonstrated significant improvements in Mean Time to Process (MTP) by 57.8% to 60%.
• Median delay was reduced by 26% to 60%, with optimized energy solutions.
• Achieved high performance metrics: precision score of 0.86, recall score of 0.82, and RMSE of 0.21.

Conclusions:

• LLM-AS effectively optimizes IoT network performance by intelligently scheduling sensing tasks.
• The proposed scheme enhances IoT usability and robustness in dynamic environments.
• LLM-based adaptive scheduling offers a promising approach for future IoT system development.