BERTHA – Bringing Human Behaviour into Automated Driving Validation

As Europe advances towards higher levels of Connected, Cooperative and Automated Mobility (CCAM), one challenge continues to limit the transition from controlled demonstrations to large-scale deployment: understanding how automated vehicles will interact with human road users in real traffic environments.

While significant progress has been achieved in vehicle perception, connectivity and artificial intelligence, automated driving systems still struggle to represent the diversity, uncertainty and complexity of human behaviour. Drivers continuously adapt their actions according to traffic conditions, perceived risks, personal preferences, cultural influences and interactions with other road users. Capturing these behavioural aspects is becoming increasingly important for the development, testing and validation of trustworthy automated mobility solutions.

The Horizon Europe project BERTHA is addressing this challenge through the development of a new generation of Driver Behavioural Models capable of reproducing key aspects of human driving behaviour and integrating them into simulation and validation environments. By incorporating behavioural realism into the development process, BERTHA contributes to more representative testing methodologies and supports the creation of automated vehicles that can interact more naturally with people.

Building the foundations for behavioural realism

One of BERTHA’s most significant achievements has been the creation of a comprehensive framework capable of representing multiple dimensions of human driving behaviour. Rather than focusing solely on vehicle dynamics or isolated driving actions, the project has explored how drivers perceive their surroundings, interpret situations, assess risk and execute manoeuvres under different conditions.

To support this objective, BERTHA analysed behavioural information from approximately 4,700 drivers, leading to the identification of eight distinct driving behaviour archetypes. This work provides a data-driven basis for understanding how different individuals respond to traffic situations and contributes to the creation of behavioural models that better reflect real-world driving diversity.

The resulting Driver Behavioural Model combines perception, cognition, affective states and motor-control responses within a modular and scalable architecture, enabling a more realistic representation of human decision-making processes.

Enabling next-generation simulation and validation

As CCAM technologies become increasingly sophisticated, validation processes must evolve to ensure that automated systems can safely operate in mixed-traffic environments.

A key outcome of BERTHA has been the deployment of behavioural modelling components through the BERTHA Data HUB and their integration into the CARLA simulation environment. These resources provide researchers, developers and industry stakeholders with access to behavioural models that can be incorporated into testing activities, virtual validation processes and automated driving development workflows.

The availability of these tools enables automated driving functions to be assessed against realistic driver behaviour representations rather than simplified traffic assumptions. This contributes to more robust system evaluation and supports the development of automated vehicles capable of responding appropriately to a broader range of traffic situations.

Supporting safer and more trusted automated mobility

Beyond technical validation, behavioural modelling plays a growing role in addressing public acceptance challenges associated with automated mobility.

Road users expect automated vehicles to behave in a manner that is understandable, predictable and socially acceptable. Systems that fail to interpret human behaviour or react unexpectedly may undermine confidence in automation, regardless of their technical performance.

By bringing behavioural intelligence into the development and validation process, BERTHA contributes to creating automated mobility solutions that are better aligned with human expectations and traffic realities. The project’s work demonstrates how behavioural modelling can complement advances in sensing, artificial intelligence and connectivity to support safer interactions between automated vehicles, conventional vehicles and vulnerable road users.

Figure 1. Integrating behavioural intelligence into automated driving enables safer, more predictable interactions with surrounding road users. Designed by Magnific.

Towards human-centric CCAM deployment

As Europe prepares for the next generation of CCAM services, understanding human behaviour will become increasingly important for both technology development and certification processes.

Through its behavioural datasets, Driver Behavioural Model, open simulation resources and validation activities, BERTHA is helping establish the foundations for automated mobility systems that are not only technically capable but also designed to operate effectively within human-centred transport environments.

The project’s results highlight the growing importance of behavioural realism as a component of future automated driving validation methodologies and represent an important step towards safer, more trustworthy and widely accepted automated mobility.

Looking Ahead

As BERTHA approaches its final phase, the project leaves behind more than a Driver Behavioural Model. It delivers a new perspective on how human behaviour can be integrated into the development and validation of automated driving systems.

Through behavioural datasets, driving archetypes, open simulation tools and human-centric validation approaches, BERTHA has contributed to strengthening the foundations for safer, more predictable and more trustworthy CCAM solutions.

As automated mobility moves closer to large-scale deployment, understanding how people perceive, decide and act will become as important as advances in sensing, connectivity and artificial intelligence. BERTHA’s results demonstrate that bringing behavioural realism into automated driving is not only possible, but essential for building mobility systems that can earn public trust and operate successfully in real-world environments.