Assessment of AGV Safety Perception in Smart Warehouse

The safety perception capabilities of Automated Guided Vehicles (AGVs) in dynamic warehouse environments are critical for collision-free operation. Existing safety assessment, however, rely heavily on fragmented manufacturer self-certification and lack standardized testing protocols, particularly for non-contact obstacle avoidance. To address this gap, this study proposes a laser point cloud-based framework for systematically assessing AGV safety perception performance across three dimensions: motion interaction, obstacle reflectivity, and spatial angle and distance perception. The methodology integrates calibrated laser point clouds with quantitative metrics to ensure reproducibility and industrial relevance. Experimental results confirmed the system’s compliance with safety application requirements in smart warehouse. AGV’s dynamic obstacle avoidance ability was assessed under varying speeds and motion modes, with minimum safe distance (MSD) as the key metric. The Obstacle reflectivity assessment quantified detection rates (DR) for objects with reflectivity ranging from $10 \%$ to $90 \%$, while spatial angle and distance experiments evaluated field-of-view (FOV) coverage and boundary recognition accuracy. The proposed methodology not only standardizes AGV safety evaluation but also provides a robust foundation for AGV safety certification processes.