Abstract

Safe yet efficient operation of professional service robots within logistics or production in human-robot shared environments requires a flexible human-aware navigation stack. In this manuscript, we propose the ILIAD safety stack comprising software and hardware designed to achieve safe and efficient motion specifically for industrial vehicles with nontrivial kinematics The stack integrates five interconnected layers for autonomous motion planning and control to enable short- and long-term reasoning. The use-case scenario tested requires an autonomous industrial forklift to safely navigate among pick-and-place locations during normal daily activities involving human workers. Our test-bed in the real world consists of a three-day experiment in a food distribution warehouse. The evaluation is extended in simulation with an ablation study of the impact of different layers to show both the practical and the performance-related impact. The experimental results show a safer and more legible robot when humans are nearby with a trade-off in task efficiency, and that not all layers have the same degree of impact in the system.