Fault-Tolerant Control of AGVs via Deep Feature Enhancement and Multi-Source Verification in Complex Industrial...

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Industrial Engineering

Background:

• Automated guided vehicles (AGVs) in intelligent material handling face challenges with environmental changes like lighting and obstacles.
• These issues cause positioning errors, mapping anomalies, and frequent, inefficient false stops.
• Existing systems lack robust perception and adaptive control for complex industrial environments.

Purpose of the Study:

• To develop a robust perception and anti-false-stop system for 2D laser-guided AGVs.
• To enhance AGV operational reliability and task handling efficiency in dynamic industrial settings.
• To address the limitations of current AGV systems in complex, real-world scenarios.

Main Methods:

• Designed a lightweight, multi-dimensional perception and anti-false-stop YOLOv8 anomaly recognition network.
• Proposed an adaptive decision-making fault-tolerant control algorithm with temporal logic verification and dynamic threshold adjustment.
• Constructed a specialized AGV anomaly detection dataset for complex industrial environments.

Main Results:

• The YOLOv8 network accurately identifies interferences in complex environments.
• The adaptive control algorithm enables real-time dynamic switching of obstacle avoidance levels.
• Real-world deployment in an electronics factory demonstrated a significant reduction in the AGV false-stop rate.
• Task handling efficiency was notably improved.

Conclusions:

• The developed system effectively solves the robust perception problem for AGVs in complex industrial environments.
• The research offers significant engineering application value by improving AGV performance and reliability.
• This approach enhances the safety and efficiency of intelligent material handling systems.