NHTSA’s March 2026 report described continuing research into how people move and are injured when seated in forward-facing, rear-facing or reclined positions inside automated vehicles. The work matters because robotaxis are designed to free passengers from the traditional upright position behind a steering wheel. A cabin with reclining seats or face-to-face passengers changes the geometry that seat belts, airbags and crash-test dummies were built to handle.
Why robotaxis change the safety problem
Conventional occupant-protection systems assume a person is sitting upright and facing forward, with the lap belt resting across the pelvis. A robotaxi may instead let a passenger recline, turn toward another passenger, work or sleep. Those positions alter the relationship between the body, restraint system and vehicle structure before a crash even begins.
That is the engineering headache in one sentence: the same belt and airbag cannot be evaluated only against the posture they were originally designed around.
US federal crash data identified 42 traffic-crash deaths over the previous decade in which the seatback was recorded as reclined or outside a normal upright position. Pre-impact posture is rarely documented, so that count does not describe the full number of such deaths.
The issue is no longer theoretical. NHTSA is studying occupant kinematics—the way the body moves during a crash—and injury mechanisms in several seating configurations. Humanetics has developed the THOR-AV 50M, a crash-test dummy intended for reclined and other nonstandard autonomous-vehicle positions.
What happens when the seat is reclined
The central risk is submarining. In a reclined posture, the occupant can slide forward beneath the lap belt instead of having the belt engage the pelvis. The restraint may then load the softer abdomen, increasing the potential for abdominal and pelvic injuries. The changed motion can also affect the lower spine, chest, legs and feet.
Airbag geometry changes too. A reclined occupant sits farther from a frontal airbag, while the legs and feet occupy a different path toward the dashboard or other interior structures. In a face-to-face cabin, the relevant impact directions and protective surfaces can differ again.
A finite-element side-impact study published in March 2023 modeled five scenarios with a WorldSID 50th-percentile male dummy: normal driving, fallback-ready, work, leisure, and relax/sleep. The most reclined scenarios produced the highest chest-compression values. The study also found that injury responses depended on the initial alignment among the dummy, the vehicle structure and the striking object.
That last point is crucial. A single crash configuration cannot represent every way a passenger might be seated. Larger seatback rotations were associated with high chest compression, and the researchers concluded that new seating positions require a broader range of impact conditions.
How engineers test nonstandard seating
Researchers combine several layers of testing rather than relying on one dramatic crash. The toolkit includes:
- finite-element simulations that model the body, restraint system and vehicle structure;
- component tests for belts, airbags and other restraint parts;
- sled tests that reproduce crash forces before a complete vehicle is built; and
- full-vehicle crashes with dummies positioned in reclined, angled or other nonstandard postures.
A staged frontal test used dummies in reclined and angled positions inside an automated vehicle. The test involved an older car traveling at 30 km/h and an automated vehicle approaching at 50 km/h, with interior footage showing how the dummies moved during the collision.
The test is useful because it makes the body-position problem visible: the dummies are not simply upright passengers viewed from a different camera angle. Their posture changes how the restraints interact with them during impact.
The dummy designed for a reclining passenger
Humanetics developed the THOR-AV 50M after NHTSA began biomechanics research covering forward-facing, rear-facing and reclined seating configurations in autumn 2018. The first prototype was assembled in mid-May 2019 after an 18-month development program.
The device modifies several parts of the conventional crash-test-dummy design:
- an updated neck with changed curvature and an added torsion element;
- a restructured upper thoracic spine and a more flexible lumbar spine;
- pelvic geometry based on anthropometric data; and
- a durable abdomen with integrated pressure sensors to help assess submarining and abdominal loading.
Those changes target the parts of the body most affected when a passenger is no longer upright. The goal is not to make a dummy look futuristic; it is to reproduce the motion and loading that standard devices may miss in a deeply reclined position.
NHTSA is studying the problem while China proposes a limit
NHTSA’s March 2026 report described continuing work on human responses, occupant motion and injuries in forward- and rear-facing reclined seating conditions. That places the US effort in the research and rulemaking phase, alongside industry testing and vehicle-specific approval processes.
China’s auto regulator proposed a 35° maximum recline for driver seats and warnings for steep recline at high speeds. The proposal is aimed at so-called “zero-gravity” seating, a term used for deeply reclined layouts designed to reduce pressure on the body during travel.
The regulatory question is broader than a single angle. Authorities must decide which postures to test, how restraints should be configured, which body sizes and ages to represent, and how a vehicle with no steering wheel or pedals should be evaluated against standards written for conventional cars.
What current robotaxis are testing
Zoox’s purpose-built robotaxi illustrates why the testing problem is different from adapting a normal sedan for automated driving. Its cabin uses face-to-face seating and has no steering wheel or brake pedals. The company describes a process that moves from simulation and component-level tests to sled tests and full-vehicle crashes.
The vehicle also uses a horseshoe-style airbag system around the passenger row, along with frontal airbags for each occupant. Those features reflect the cabin’s layout, but Zoox’s own validation program remains manufacturer testing rather than an independent safety rating for reclined occupants.
The broader shift is straightforward: once passengers can sit in positions that cars were not designed around, crash safety has to measure more than the vehicle’s structure. It has to account for where the body starts, how the belt holds it, how the airbag reaches it and how the occupant moves through the cabin.
