Autonomous vehicles are designed to make driving safer, but their interactions with human drivers introduce challenges that technology alone cannot eliminate.

A car programmed to maintain a consistent following distance may behave very differently from a human driver responding to changing traffic conditions.

A new study published in Accident Analysis & Prevention, Volume 234, examines how these differences influence driving behaviour and collision risk. Using real-world vehicle data, researchers identified important contrasts between automated and human-driven cars, particularly when one follows the other.

How Different Vehicles Interact

The researchers analysed real-world car-following data from the OpenACC database. Their investigation compared automated vehicles (AVs) and human-driven vehicles (HVs) across different driving scenarios and examined how the type of vehicle ahead affected following behaviour.

Rather than considering all traffic interactions equally, the study explored how drivers responded to specific situations, including highway driving and controlled test scenarios.

The team combined descriptive statistics with mixed-effects models to identify relationships between driving conditions, following behaviour and risk indicators.

Human Drivers Change Their Behaviour

One of the clearest findings was that human drivers adapt their behaviour considerably depending on which vehicle they follow.

On highways, their speed patterns showed two distinct peaks. In controlled test scenarios, they generally maintained longer following distances and demonstrated greater caution when travelling behind automated vehicles.

When following other human drivers, however, they tended to behave more casually and occasionally exceeded speed limits.

Automated vehicles responded differently. Their behaviour remained comparatively consistent, reflecting programmed responses to the movements of the vehicle ahead.

Why Short Distances Matter

The study revealed an important distinction between how frequently risky situations occur and how serious those situations become.

Automated vehicles experienced a higher rate of potentially risky situations, particularly those involving short following distances. However, the severity of those risks was generally lower.

They also demonstrated more consistent responses when Time-to-Collision, a measure of how quickly vehicles could collide if their relative movements continued unchanged, reached low levels.

This suggests that automated driving systems may manage certain dangerous situations proactively, even when their following behaviour produces more frequent warnings.

The Most Challenging Combination

The researchers identified particularly important differences in mixed traffic.

Human drivers faced significantly greater risk in highway scenarios and when following automated vehicles. However, the highest risk occurred when an automated vehicle followed a human-driven car.

This combination highlights the difficulty automated systems face when responding to less predictable human behaviour.

Researchers Lianzhu Sun, Xiaomei Zhao and Penghui Li found that driving scenarios and the type of leading vehicle had stronger effects on human drivers than on automated systems.

What This Means For Road Safety

The findings suggest that assessing autonomous vehicle safety requires more than comparing automated driving with human performance in isolation.

The researchers emphasised that road conditions, automation levels and the combination of vehicles travelling together all influence collision risk.

For manufacturers, understanding these interactions could help improve automated following strategies. For traffic planners, the results highlight the importance of considering mixed traffic as automated vehicles become more common.

The study does not establish that autonomous vehicles are inherently more dangerous than conventional cars. Instead, it shows that risk depends on the situation and the vehicles involved.

As automated driving expands, understanding how humans and machines respond to one another will be essential to developing safer roads during the transition towards increasingly automated transport.