How can the performance of a driver assistance system be evaluated objectively without equipping the vehicle under test with specialized measurement equipment? Experiments conducted on the M1–M7 smart road section by the BME Department of Automotive Technologies explore this question by testing ACC and Lane Centering functions in real highway traffic, using a digital twin generated in real time from infrastructure sensor data.

Testing advanced driver assistance systems (ADAS) in a reliable and meaningful way is becoming increasingly complex. Functions such as Adaptive Cruise Control (ACC) and Lane Centering do not operate in isolation: they continuously respond to a dynamic and changing traffic environment. This makes it increasingly important to complement laboratory and closed proving-ground tests with methods that allow ADAS performance to be evaluated objectively and reproducibly under real road traffic conditions.

Researchers at the BME Department of Automotive Technologies conducted a series of tests on the M1–M7 smart road section to investigate the operation of ACC and Lane Centering in real highway traffic. The distinctive feature of the measurement environment was the simultaneous generation of a digital twin of the road section from the infrastructure’s sensor data in real time. The digital twin represented not only the test vehicles, but also the surrounding traffic with high accuracy.

A total of five test vehicles participated in the measurement campaign, three of which were also equipped with high-precision GNSS systems. These vehicles provided independent reference data for validating the accuracy of the digital twin, including vehicle position, speed, relative distance and motion trajectories.

The tests were conducted in both traffic directions and across all available lanes of the M1–M7 smart road section: four lanes in each direction, eight lanes in total. A key characteristic of the experiments was that they were carried out not in an isolated test environment, but alongside actual highway traffic.

This setting also allowed the researchers to observe not only predefined test scenarios, but naturally occurring and spontaneous traffic events.

“One of the key advantages of the digital twin is that it does not only see the test vehicle, but its entire traffic environment as well. This effectively provides an external reference that makes it possible to objectively evaluate the behaviour of ADAS systems in real traffic.”
– Dr Zsolt Szalay, Head of the BME Department of Automotive Technologies

ACC Response in Controlled and Spontaneous Situations

Several traffic scenarios were developed to investigate Adaptive Cruise Control. In the initial scenarios, the lead vehicle travelled at a constant speed while the ACC system of the following vehicle regulated its speed according to the configured following conditions.

In further scenarios, the lead vehicle performed accelerations and decelerations of varying magnitude. These situations made it possible to analyse the response of the ACC system in detail, including the temporal evolution of following distance and relative speed, as well as how dynamically the system responded to changes in the speed of the vehicle ahead.

At the same time, real traffic introduced situations that could not be fully planned in advance. On several occasions, for example, a vehicle from the public traffic merged into the gap between the two test vehicles.

Although such events caused the measurements to deviate from the predefined test scenarios, they were particularly valuable from a research perspective. Such situations are much closer to the conditions that the ADAS system of a production vehicle may encounter in everyday traffic.

Because the infrastructure-based digital twin also tracks vehicles in public traffic, these spontaneous interactions can be reconstructed and analysed. This makes it possible to determine when, where and how the traffic situation changed, and how the ACC system responded to the new conditions.

Lane Centering: Making In-Lane Vehicle Motion Measurable

During the Lane Centering tests, the researchers analysed vehicle lane-keeping behaviour and in-lane trajectories. Based on the high-precision vehicle trajectories determined by the digital twin, the lateral position of the vehicle within its lane and its evolution over time can be evaluated.

This is particularly relevant for a function whose operation is primarily reflected in the continuous control of the vehicle’s lateral position. An infrastructure-based reference makes it possible to assess vehicle motion independently of the vehicle itself, using an external measurement system.

The approach therefore goes beyond simply “seeing” where a vehicle is travelling. The available data make it possible to quantify the vehicle’s position within the lane and how that position changes over time.

When Traffic Itself Becomes Part of the Test

One of the major challenges of conventional testing is that, in real road environments, other road users can change the conditions in unpredictable ways. In the present measurements, however, this characteristic of real traffic became one of the most interesting research opportunities.

In addition to the test vehicles, the digital twin represented the surrounding public traffic. As a result, an unexpected lane change, a change in speed or another traffic interaction was not simply an “unplanned event”, but became a measurable and analysable occurrence.

“Spontaneous interactions in real traffic do not necessarily disrupt a test – in many cases, they can provide valuable data. With the help of the digital twin, we can also analyse situations that develop unexpectedly during test execution.”
– Dr András Rövid, Head of the Cooperative Perception Research Group

This approach can be particularly important for evaluating highly complex ADAS functions, where system behaviour is often determined by interactions involving multiple vehicles and continuously changing traffic situations.

Digital Twin as a Vehicle-Independent Measurement Reference

One of the key objectives of the experiments was to demonstrate that a digital twin generated in real time from infrastructure sensor data can be used to evaluate ADAS functions against an external, vehicle-independent reference.

The three GNSS-equipped test vehicles provided high-precision reference data that enabled the researchers to verify how accurately the digital twin could determine vehicle positions, speeds, relative distances and trajectories.

The significance of this approach is that, in the longer term, it may eliminate the need to equip every vehicle under test with specialized high-precision GNSS measurement systems in order to obtain objective measurement data.

If infrastructure can reconstruct the motion of traffic participants with sufficient accuracy and in real time, the road itself – more precisely, its intelligent sensor network and the digital twin generated from it – can serve as an external measurement reference system.

This opens up new possibilities for testing ADAS systems on a larger number of vehicles in real road environments. The approach makes it possible to analyse the motion of multiple vehicles simultaneously, the interactions between them, and the responses of driver assistance systems to these interactions.

One Step Closer to Scalable ADAS Testing

The experiments conducted on the M1–M7 smart road section therefore represent more than the evaluation of individual ACC and Lane Centering functions. They also demonstrate a testing and evaluation concept in which real-time infrastructure sensor data, a digital twin and the operation of vehicle-based driver assistance functions are integrated into a common system.

One of the most promising aspects of this approach is its potential to reduce the need for specialized vehicle-side measurement equipment, while enabling the collection of large volumes of objective and comparable test data.

This could become particularly important as ADAS functions and automated driving systems continue to evolve and increasingly complex real-world traffic situations need to be measured, evaluated and compared.

From this perspective, the M1–M7 smart road section is more than an intelligent road. It provides a research environment in which the sensing capabilities of road infrastructure can be directly connected to the objective evaluation of vehicle behaviour – bringing the testing of ADAS functions one step closer to scalable, infrastructure-based validation in real traffic.