Q&A: Key Facts About Aptiv’s Advanced Occupancy Classification

To ensure the best possible safety, seatbelt tensioning and airbag deployment should be tailored to each occupant in a vehicle. It is not enough to know which seats people are sitting in; the system should know their heights, weights, postures and seating positions. Traditional, bladder-based occupation detection is reliable, but it cannot make such fine-grained distinctions. To achieve a fuller spectrum of safety, Aptiv uses the interior camera typically required for global safety regulations to perform occupant classification as well. Here, Timo Rehfeld, engineering manager for AI/ML interior sensing at Aptiv, discusses the details of this innovative approach.

What is Advanced Occupancy Classification?

AOC is the industry’s first occupant-detection system powered entirely by software and a vehicle’s interior camera, replacing bladder, weight or capacitive seat sensors inside the seat cushion. The name is intended to distinguish it from those other methods, which can detect that a seat is occupied but can’t classify the occupant’s personal characteristics, which are critical factors for proper airbag deployment force, seatbelt tensioning and restraint strategies. Put another way, we all know there’s a difference between a 6-foot-tall man, a 5-foot-tall woman and a 3-foot-tall child. When a vehicle knows it too, it can automatically and immediately make lifesaving adjustments.

What is the most surprising benefit of AOC?

AOC allows for far more flexibility in interior cabin design. In the short term, that can mean more comfortable cabins, with the ability to easily reconfigure seats — for example, when groceries, luggage or assistive devices need to be placed in the vehicle. In the future, when the goal of fully autonomous driving is achieved, it will mean the entire cabin can be rearranged, either by OEMs or consumers, to be anything from a mobile office to an entertainment center on wheels.

Will OEMs need additional inputs or sensors beyond the interior camera?

No. AOC only requires the vehicle’s interior camera, which is already present in most vehicles for driver monitoring or cabin monitoring. It works by employing trained vision models to analyze camera frames, followed by temporal tracking across frames and rule-based logic to improve robustness and decision consistency.

The intent of the design is to keep costs low by taking advantage of what already exists in most vehicles. By leveraging existing hardware and boosting its abilities with AI and machine learning software, the system avoids the need for dedicated in-seat sensors while maintaining high accuracy. And basing the capabilities on software means that upgrades can be handled over the air.

How much can AOC reduce the cost of implementing an occupant detection system?

Because AOC does not need dedicated seat-based occupancy hardware, we estimate the savings for OEMs could be as high as 40 percent. The exact number will depend on factors such as the vehicle’s baseline architecture, the cost of the displaced hardware and platform scale.

To which regulatory requirements does AOC conform?

AOC meets global regulatory requirements, including FMVSS 208 requirements for airbag suppression and low-risk deployment, demonstrating 100 percent accuracy across all regulatory test cases. The system achieves over 99 percent accuracy in real-world “due care” scenarios, reflecting consistent performance under everyday conditions. AOC is also aligned with Euro NCAP protocols, enabling OEMs to support high safety ratings across global markets. This ensures that AOC delivers both regulatory compliance and strong performance in consumer-facing safety assessments.

It’s also ready for future regulations that may require more detailed occupant classification, such as distinguishing additional occupant groups or identifying rear‑facing versus forward‑facing child seats.

How has the system been tested?

Aptiv performed nearly 5,000 tests, from formal regulatory tests to a wide range of possible scenarios. The testing included occupants of different ages, sizes and seating behaviors. We made sure to test conditions that could challenge a sensor — such as bulky clothing, occlusions and various kinds of child seats — to ensure that AOC could perform reliably in the real world, not just in controlled environments.

How does AOC perform if camera visibility is temporarily compromised?

When visibility is compromised, AOC automatically defaults to conservative, regulation-compliant behavior and quickly restores full performance once conditions improve, ensuring safe and reliable operation. It also uses vehicle signals — such as seat belt buckle sensors, door status sensors and vehicle motion data — for detecting occupancy in periods of degraded visibility. AOC can detect both camera blockage and common forms of occlusion. For example, it can identify situations where the camera is covered by a sticker or when an occupant or a child seat is obscured by a blanket. In these cases, AOC transitions to a safe state and alerts the user.

Is AOC in production?

Yes. We are already in preliminary production with a global OEM, with full ramp-up scheduled for 2027. As part of a phased rollout, the OEM will initially deploy both traditional, seat-based systems and camera-based AOC in parallel, transitioning to a fully camera-based, software-only system by 2029.