Overview
A large-scale autonomous mobile fulfillment robot was developed for transporting heavy payloads within active warehouse environments. The system operates in human-populated spaces and is designed for high reliability and high availability during navigation, including high-speed transit while carrying payloads. A multi-modal sensor suite combining LiDAR, time-of-flight (ToF), and ultrasonic sensors is used to support obstacle detection, navigation, and collision avoidance. The safety concept integrates non-SIL sensors within a SIL2-oriented system architecture. 

Challenges
The primary challenge was ensuring safety and reliability in a dynamic warehouse environment using a diverse sensor set that included non-SIL-rated components within a system targeting SIL2 safety level. This required a different system-level safety design to accommodate heterogeneous sensor reliability characteristics. 

Additional complexity arose from: 

• Blind-spot risks caused by payload-induced occlusion 

• Differentiation between dynamic obstacles (e.g., humans, moving equipment) and static infrastructure 

• Maintaining safe operation under varying environmental and operational conditions in a continuously changing workspace 

Considerations included balancing safety, performance, and system availability: 

• High sensitivity for obstacle detection improved safety but increased unnecessary stops and reduced throughput 

• More permissive detection thresholds improved operational efficiency but increased risk of missed or delayed hazard detection 

• Sensor diversity improved redundancy but increased system complexity and calibration effort 

• High-speed operation under payload conditions required careful tuning of detection latency versus robustness 

Approach
A structured safety architecture was developed around multi-layered perception and monitoring: 

• Use of diverse sensor modalities (LiDAR, ToF, ultrasonic) to provide redundant environmental perception 

• Integration of non-SIL sensors within a SIL2 safety concept through system-level redundancy and validation logic 

• Implementation of safety monitor and watchdog architectures to supervise perception and control behavior 

• Definition of Operational Design Domain (ODD) boundaries, including environmental constraints and operational limits 

• Development of safe-degraded modes to maintain controlled behavior under sensor uncertainty or failure conditions 

• Blind-spot management strategies to mitigate occlusion effects from payload geometry 

• Dynamic classification logic to distinguish between moving obstacles and static infrastructure elements 

Key Insights
The project highlighted several important findings:

• Safety in autonomous warehouse systems depends on system-level architecture rather than individual sensor safety ratings 

• Non-SIL sensors can be effectively used in safety-critical systems when supported by appropriate redundancy and monitoring layers 

• Blind-spot and occlusion effects become primary safety drivers in payload-carrying mobile robots 

• Clear ODD definition is essential to ensure predictable and certifiable system behavior under real operational conditions 

• Safe-degraded modes are critical for maintaining operational continuity without compromising safety integrity 

Outcome
A robust safety architecture was developed enabling reliable deployment of autonomous mobile fulfillment robots in high-density warehouse environments. The system achieved improved collision avoidance performance while maintaining high operational availability. Through structured use of multi-modal sensing, safety monitoring, and degraded-mode behavior, the platform successfully balanced safety integrity with throughput requirements in real-world warehouse operations.