Camera-in-the-Loop in ADAS Simulation: integration, validation, and use cases
At the Smart Prototypes Summit 2026, Cristina Petrucci from Automobili Lamborghini and Claudio Annicchiarico from Meccanica 42, in collaboration with VI-grade, presented an advanced Camera-in-the-Loop ADAS simulator for super sports cars.
The solution integrates a real camera sensor within a virtual driving environment, enabling closed-loop validation of perception algorithms under realistic and repeatable conditions.
The objective is to reduce the gap between simulation and on-road testing while increasing scenario coverage, reproducibility, and development efficiency.
ADAS development in super sports cars: system-level constraints
For super sports car manufacturers, ADAS systems represent a complex trade-off driven by a fundamental dual requirement:
- compliance with safety and homologation regulations: the implementation of a minimum set of safety functions is required to ensure homologation and market access.
- preserving an uncompromised driving experience: these functions must remain completely transparent to the driver: any false positive, false negative, or unnecessary intervention can degrade driving experience and compromise vehicle identity.
From a vehicle integration perspective, these constraints are amplified by architecture complexity. Automobili Lamborghini’s cars must accommodate up to 23 ADAS sensors within geometries characterized by low ride height, wide track, and highly inclined windshields.
These conditions impose strict limitations on both sensor packaging and mounting feasibility, as well as on the calibration of system functions, which must simultaneously meet performance targets and homologation requirements.
Beyond initial integration, ADAS development requires continuous system-level revalidation. Every software update introduced by partner ECUs triggers full regression testing across the complete functional chain.
When combined with limited prototype availability and the execution of many maneuvers across extensive mileage and heterogeneous scenarios (such as varying lighting conditions, traffic environments, and edge cases), this significantly reduces the throughput of physical validation activities.
Considering that super sports car annual production volumes are limited, the cost associated with ADAS development and validation is difficult to amortize across the fleet, making extensive on-road testing economically inefficient.
The role of HiL in ADAS validation
In this context HiL-based approaches have become an essential tool for ADAS development, supporting not only functional safety and SOTIF activities but also system-level and vehicle-level validation.
HiL methodologies are increasingly aligned with regulatory trends, including UNECE frameworks such as R157 and emerging validation methodologies based on multi-pillar approaches to robustness and repeatability.
The key advantage of HiL lies in enabling validation of real components within a simulated environment, while keeping the physical prototype limited to selected subsystems, improving efficiency, repeatability, and scalability of the development process.
Camera-in-the-Loop implementation
That’s why Automobili Lamborghini introduced Meccanica 42’s CamiL into the ADAS development workflow as a strategic enabler, supporting:
- system-level and vehicle-level validation
- controlled scenario reproduction
- repeatable testing conditions
- reduction of physical prototype dependency.
CamiL integrates the physical camera sensor into the simulation loop, allowing:
- closed-loop testing of ADAS functions
- actuator feedback (including AEB braking)
- real-time interaction between virtual environment and physical sensor.
The Camera-in-the-Loop test rig was developed through an iterative roadmap, progressively increasing system realism and integration depth.
Initial validation focused on FlexRay communication, using a minimal configuration of three real ECUs, while the remaining architecture was fully virtualized. This enabled early verification of signal consistency in a hybrid virtual-physical setup.
The next step involved integrating the camera with the virtual vehicle architecture. This required analysing and virtualising the communication network to reproduce the interactions between the ADAS camera and the rest of the vehicle.
While a modern vehicle may exchange around 600 signals across its ECUs, only about 20% are required for Camera-in-the-Loop integration, as many are related to diagnostics or non-perception functions. Selecting and virtualising only the relevant signals simplifies integration and reduces system complexity.
To ensure realistic sensor behavior, once the camera is integrated into the simulation loop, it is calibrated using procedures equivalent to those performed in end-of-line production stations, including adjustment of focal length, sensor position, brightness, and field of view.
Then, the vehicle dynamics model is introduced, marking the first activation of ADAS functions within the simulator environment.
Once the platform reached a stable configuration, the simulator incorporates both the driver station and the production infotainment system, allowing interaction with ADAS functions through the same interfaces available in the vehicle.
Validation of perception performance
A key objective of the project was to validate the perception performance of the camera system in a realistic and measurable way.
Perception performance was evaluated using the distance-to-line-crossing metric from the lane departure warning function.
This parameter was used as reference to compare:
- ECU output in real system conditions,
- equivalent computation derived from virtual scenarios.
Results showed strong correlation and consistent dynamic behavior between the two datasets. A small constant offset was observed, attributable to differences in vehicle width modeling: the virtual model includes side mirror contribution, which is not considered in the ECU calculation.
Overall, results confirm that the Camera-in-the-Loop environment provides a reliable and repeatable framework for ADAS perception validation under closed-loop conditions.
Conclusion
As ADAS functions continue to grow in complexity and validation requirements expand, Camera-in-the-Loop and Hardware-in-the-Loop approaches are becoming an essential part of the development process.
By combining real sensors with virtual environments, manufacturers can increase validation coverage while reducing dependence on costly and time-consuming road testing.
This is particularly relevant for low-volume manufacturers, where regulatory-compliant ADAS systems must be implemented across a limited number of vehicles, making cost-efficient validation approaches a critical enabler of sustainable development processes.
For more information about our innovative solutions and how they can streamline your vehicle’s functions development, please contact us.
Don’t forget to follow us on LinkedIn and Instagram, and to subscribe to our newsletter to stay in the loop!