INTEGRATING EDGE AI VISION FOR RELIABLE PATH TRACKING IN AUTOMATED GUIDED VEHICLES

Abstract

Conventional automated guided vehicles (AGVs) frequently lack adaptability in dynamic Industry 4.0 environments due to their reliance on basic sensor arrays and centralized processing. This study investigates the enhancement of an AGV through the integration of an Edge Artificial Intelligence (AI) vision module, leveraging decentralized Edge AI for real-time on-device inference to eliminate centralized processing latency, to evaluate its operational performance and reliability. A baseline AGV utilizing infrared line sensors and ultrasonic obstacle detection was empirically compared against an enhanced AGV equipped with a HuskyLens vision module on an identical chassis. Systematic evaluations were conducted across multiple performance metrics under controlled conditions. The experimental results indicate that both systems exhibited comparable reliability in line-tracking success (95% for the baseline versus 91% for the AI-enhanced AGV) and obstacle detection rates (90% versus 93%, respectively). However, the integration of Edge AI introduced operational trade-offs, evidenced by an increased mean lap travel time (20.46 s compared to 16.38 s) and a reduced battery endurance (31.3 min compared to 43.6 min) owing to higher computational demands. Despite these constraints, the enhanced AGV successfully demonstrated advanced capabilities, including object classification and face recognition, highlighting the potential of Edge AI for executing complex autonomous tasks.