Automating ADAS/AD function validation: A Smarter, Safer Path to Autonomous Mobility

In the race toward safer, smarter mobility, Advanced Driver Assistance Systems (ADAS) and Autonomous Driving (AD) technologies are no longer futuristic—they’re foundational. However, as these systems become increasingly complex, the challenge of validating them also grows. Traditional road testing is expensive, not repeatable, and time-consuming. So, how do we ensure these systems are safe, reliable, and ready for the real world?

As recently discussed at Safety.AD and in a blog post from ESTECO (an independent software company specializing in digital engineering), the importance of testing new ADAS and AD functions remains despite reduced development times and pressures to streamline costs. In fact, it has become increasingly imperative that the industry considers new ways to align software simulations with real-world road testing.

To address this, ESTECO, in collaboration with IPG Automotive, Keysight, and TNO, has developed a structured and automated approach to ADAS/AD function validation. This approach integrates simulation, hardware-in-the-loop (HIL) testing, and probabilistic modeling to create a robust and scalable validation pipeline. It aligns closely with the industry’s growing emphasis on early fault detection through simulation.

The Problem: Complexity, Cost, and Coverage

ADAS/AD functions must operate safely across a vast range of Operational Design Domains (ODDs)—from urban traffic to rural highways. Validating these systems means accounting for:

Seemingly infinite scenario permutations
Post-production updates in Software-Defined Vehicles (SDVs)
Evolving algorithms and sensor behaviors
Compliance with SOTIF (ISO 21448) and FuSa (ISO 26262) standards

Road testing alone can’t keep up. It’s costly, non-repeatable, and limited in scope. Enter simulation. The complexity of ADAS/AD validation stems from the need to ensure safe operation across a wide array of Operational Design Domains (ODDs). These domains encompass everything from urban intersections to rural highways, each presenting unique challenges. Moreover, the rise of Software-Defined Vehicles (SDVs) introduces additional layers of complexity, as post-production updates and evolving algorithms demand continuous validation. Standards such as SOTIF (Safety of the Intended Functionality) and ISO 26262 further underscore the importance of rigorous safety assurance.

Historically, Tier 1 suppliers and OEMs have used X-in-the-loop methodologies—starting with software-in-the-loop (SIL) and gradually integrating real hardware—to validate vehicle systems. While effective for traditional domains like powertrain and chassis, ADAS introduces new challenges. Sensor behavior must be modeled with high fidelity, and the distributed nature of ADAS functions demands tighter integration across test systems.

The Solution: A Multi-Partner Validation Pipeline

The experiment discussed at Safety.AD showcased the complete workflow from software simulation to in-lab scene generation, road testing and compliance with standards that will mitigate the safety risks.

Step 1: Scenario Generation with modeFRONTIER + CarMaker

Using modeFRONTIER, ESTECO automates the generation of test scenarios in CarMaker (IPG Automotive’s simulation solution). Parameters like vehicle speed, overlap, acceleration, and interdistance are varied across statistically distributed values derived from real-world driving data.

This creates a data lake of synthetic scenarios that reflect realistic driving conditions without ever leaving the lab.

ESTECO’s solution begins with scenario generation using modeFRONTIER, which automates the creation of test cases within IPG Automotive’s simulation solution CarMaker. This setup then serves as the foundation for evaluating system performance under diverse conditions.

Step 2: Realistic Radar Emulation with Keysight RSE

Simulating radar behavior is notoriously tricky. Proprietary filtering algorithms and environmental noise make it hard to replicate real-world sensor responses.

Keysight’s Radar Scene Emulator (RSE) solves this by using rixels—RF transceivers that emulate radar echoes in high resolution. Integrated with CarMaker, the RSE enables hardware-in-the-loop testing, by generating highly realistic radar echoes from the simulation environment and allowing the actual radar hardware to respond as it would in real-world conditions.

Taking this approach further leads to vehicle-in-the-loop (VIL) testing, which combines the repeatability and automation of lab-based setups with the realism of full vehicle integration. In both HIL and VIL configurations, the radar sensor is placed inside a sealed chamber, isolated from external disturbances, and synchronized with the simulated environment. This setup allows for continuous, reproducible testing that can be conducted 24/7, independent of weather or track availability. For more information, you can read this article by Keysight's Henrik Liebau.

Step 3: Probabilistic Modeling with TNO

Once the data is collected, TNO steps in to build a physics-informed probabilistic model. This model:

Assesses detection performance, positioning accuracy, and sizing
Identifies failure-contributing factors
Enables both forward reasoning (predicting performance) and backward reasoning (exploring edge cases)

In one case study, the presence of a guardrail was shown to obscure radar detection, leading to potential false negatives. TNO’s model quantified this risk, demonstrating how environmental factors can compromise safety. Once the simulation and emulation data are collected, TNO applies a physics-informed probabilistic model to analyze system reliability. This model evaluates detection performance, positioning accuracy, and object sizing, while also identifying factors that contribute to system failures.

A screenshot of a computer AI-generated content may be incorrect.

The Impact: Efficiency, Transparency, and Safety

This automated pipeline delivers several key advantages across the development lifecycle:

Shorter development cycles through front-loaded virtual testing
Objective risk quantification to support certification and compliance
Optimized resource usage, reducing reliance on physical prototypes and road testing
Transparent traceability of results to meet regulatory requirements
Enhanced sustainability by minimizing environmental impact

Most importantly, the system enables continuous, automated testing—24/7 and independent of weather or track availability. By shifting testing earlier in the process and reducing physical dependencies, teams can iterate faster, make data-driven decisions, and validate more thoroughly.

What’s Next?

Looking ahead, the team plans to expand the probabilistic model to include additional insufficiencies such as reflections from road surfaces and object fusion during complex maneuvers. They also aim to incorporate real vehicle dynamics and actuation responses, automate the optimization loop to identify critical test cases, and quantify vehicle risk performance more precisely. The ultimate goal is to create an enterprise-ready environment where each partner can contribute while maintaining intellectual property protection and ensuring data traceability.

In conclusion, the collaboration between ESTECO, IPG Automotive, Keysight, and TNO offers a powerful solution for virtual testing and validation of ADAS/AD functions. By combining simulation, emulation, and probabilistic modeling, this approach enables automotive OEMs and Tier 1 suppliers to develop safer, more reliable systems while accelerating time to market and reducing costs. This isn’t just about automation—it’s about building trust in the future of autonomous mobility.