Overview
A high-volume sorting system was developed for handling elevated items potentially exceeding 50 lbs in a shared work environment with human operators. The system supports both manual loading of the cell and fully automated loading with no human in the loop. Therefore, it is required to operate safely in close proximity to humans while also maintaining fault-tolerant operation during automated loading cycles without intermittent failures. A key safety requirement is that items must never be dropped onto a human, regardless of weight or operating mode. The system also needed to be robust against common forms of unintentional operator misuse. This case highlights application-specific sensor selection and safety function design in complex human–machine environments. 

Challenges
The primary challenge was achieving a cost-effective yet reliable method to distinguish between human presence and portable storage receptacles used within the work cell. Conventional safety-rated sensing solutions were considered cost-prohibitive and not optimally suited for the application. 

Additional complexity arose because triggering safety functions based on receptacle placement was not viable, and accidental activation of sorting behavior due to human presence had to be robustly prevented. The resulting work-cell logic required behavior beyond what is typically addressed in IEC 61496-based guidance. 

Considerations were centered around the similarity between human operators and storage receptacles in terms of physical footprint and placement within the workspace: 

• If receptacle acceptance criteria were too strict, operators would be incentivized to bypass safety mechanisms, reducing overall system safety 

• If criteria were too permissive, the system risked misclassifying a human as a receptacle and initiating hazardous sorting operations 

Balancing operational efficiency and safety integrity required careful tuning of detection and classification logic. 

Approach
A structured safety engineering approach was used to define and constrain system behavior within the intended Operational Design Domain (ODD): 

• Definition of ODD boundaries including human variability (height, posture, clothing, behavior, and intent), explicitly rejecting assumptions such as “humans will never be present” or “specific clothing types will never occur” 

• Development of sensor selection and placement strategies to enable reliable discrimination between humans and receptacles under real-world variability 

• Design of internal system logic supporting safe operation for personnel manually feeding the cell while minimizing nuisance triggering during fully automated robotic feeding. 

• Integration of safety functions that preserve operational flexibility without relying on overly restrictive or overly permissive classification rules 

• Development of safety arguments aligned with the intent of consensus standards, while extending beyond strict clause-level interpretation where required for the application 

Key Insights
The project highlighted several important findings:

• A defined ODD enables more defensible and practical safety architecture decisions 

• Human variability (size, posture, clothing, intent) must be explicitly modeled in industrial safety systems rather than assumed away 

• Presence and absence logic used for human detection in safety systems should be analyzed to a much higher degree if the outcome of the detection logic could result in machine enabling functionality 

• Effective safety engineering may require interpretation of consensus standards beyond literal compliance to achieve intended safety outcomes in complex applications 

Outcome
The work cells were successfully retrofitted with additional safety functions that maintained operational performance while significantly improving personnel safety. Sensor selection, placement, and system logic were optimized to accommodate variability in storage receptacles while ensuring robust human protection. 

Retrofit costs and operational disruption were minimized, enabling scalable deployment across both low-automation and high-automation sites. The resulting safety solution achieved broad adoption and demonstrated improved safety performance without significant reduction in throughput.