From Virtual Testing to Public Roads: How MOIA Validates Autonomous Vehicles

Before our autonomous vehicle, equipped with Mobileye self-driving technology, can begin operating as part of an on-demand mobility service, it must successfully complete a comprehensive testing and validation process. At MOIA, that process spans virtual simulations, hardware based testing, closed-course evaluations, and public road operations. Each stage provides essential evidence that helps demonstrate the safety, reliability, and performance of the autonomous driving system.

Testing is fundamental to understanding how an autonomous system performs under real operating conditions.

"Comprehensive testing enables us to demonstrate that our autonomous vehicles meet the requirements for safe operation on public roads," says Dr. Lars Gehrke, SVP Autonomous Driving Test Operations.

Autonomous vehicles must perform reliably under far more than ideal driving conditions. They need to operate safely across a broad range of traffic situations, weather conditions, and unexpected events encountered in everyday traffic. No single test can provide sufficient evidence. Instead, confidence is built through the combination of multiple testing methods and thousands of variations of carefully designed test scenarios.

Each testing environment serves a distinct purpose. Together, they form an integrated validation process that supports the safe development of autonomous mobility.

The validation process begins entirely in a virtual environment.

Dr. Christina Bogan, Head of Virtual Testing, and her team evaluate the digital driver across millions of simulated traffic scenarios.

"The digital driver is the software that controls the vehicle instead of a human driver."

Virtual simulations verify that the autonomous driving software performs as intended. Engineers assess whether the system correctly interprets traffic rules, recognizes road users, and responds appropriately to a wide variety of driving situations. This stage focuses on verification, demonstrating that the automated driving function behaves according to its design specifications.

Simulation provides a scalable and highly controlled environment for evaluating autonomous driving systems long before testing begins with physical vehicles. Compared with hardware based testing or public road operations, Software in the Loop (SiL) simulations offer several important advantages:

• Scalability and automation. Thousands of simulations can run simultaneously without continuous human supervision, allowing engineers to evaluate a significantly larger number of scenarios than would be possible with physical testing alone.
• Flexible scenario generation. Traffic situations, road layouts, weather conditions, lighting, and countless environmental variables can be modified efficiently to create diverse testing conditions.
• Risk free testing. Complex and potentially hazardous situations can be evaluated safely without exposing people, vehicles, or equipment to unnecessary risk.

In addition to virtual simulations, Hardware in the Loop (HiL) test benches play a critical role in verifying autonomous mobility systems. These test environments combine physical vehicle components, such as the steering system or turn signals, with the digital driver's computing hardware, sensors, and software in a simulated environment.

Here the autonomous mobility service is evaluated as an integrated ecosystem. In addition to the vehicle itself, the focus is on the cloud-connected functions of the mobility service. The test benches are used to verify and analyze in detail how all system components work together.

"During end-to-end testing, we evaluate the entire customer journey. The focus is on the complete mobility system and on understanding how the vehicle, the digital driver, and the cloud-based mobility services interact as one integrated solution," explains Dr. Christian Rösener, Head of Verification & Simulation.

On the test bench, the entire journey is evaluated – from booking the vehicle and its arrival, to boarding, including the safety check, through to the ride itself, arrival at the destination, and exiting the vehicle.

Testing the complete system under controlled conditions allows engineers to evaluate complex interactions long before they are exposed to real traffic. This approach accelerates development while eliminating risks to people and physical assets.

Testing on a closed test track is another key step in the verification and validation of autonomous driving systems.

At the test track in Munich, MOIA evaluates how the autonomous vehicle responds in predefined as well as particularly complex traffic scenarios, helping to verify the safety and reliability of the overall system.

"The proving ground bridges the gap between simulation and real-world driving. It allows scenarios to be recreated that would be difficult or unsafe to evaluate in everyday traffic," explains Lars, SVP Autonomous Driving Test Operations.

Typical proving ground scenarios include:

• pedestrians, vehicles, or unexpected objects suddenly entering the roadway 
• disabled vehicles or emergency response situations 
• unusual driving behavior by other road users, such as abrupt lane changes or unexpected stops 
• complex intersections and turning maneuvers involving multiple road users 

The proving ground also serves as the primary training environment for MOIA's safety drivers. Here, they become familiar with the autonomous driving system across a broad range of operating conditions while supporting the collection of valuable operational insights.

The experience gained during these exercises is continuously incorporated into the further development of the autonomous driving system.

While simulations and controlled test environments are an essential part of development, autonomous driving technology must ultimately demonstrate its performance under real-world traffic conditions.

MOIA’s public road testing demonstrates how autonomous vehicles handle the wide variety of real-world traffic situations and interact with other road users in everyday operation. The autonomous ID. Buzz is already being tested in seven cities worldwide, including Hamburg, Berlin, Oslo, Munich, Austin, Los Angeles, and Orlando.

Throughout every test drive, a trained safety driver remains behind the wheel. Although the vehicle operates autonomously, the safety driver continuously monitors system performance and is prepared to intervene whenever necessary.

"Rare and therefore particularly valuable situations are automatically captured by the vehicle, while our safety drivers provide additional context through their observations. These insights are fed directly back into development, enabling continuous improvements to both the digital driver and the overall mobility system," says Lars.

An essential component of public road testing is MOIA's continuous feedback process:

• Safety drivers identify noteworthy or unusual situations during testing. 
• Vehicle data is recorded throughout every trip. 
• Development teams analyze system behavior and continuously refine the autonomous driving software. 

This ongoing feedback loop enables every test drive to contribute directly to the continuous improvement of the autonomous mobility system.

All testing phases contribute to three core objectives:

Verification:
Does the system meet the defined requirements?

Validation:
Is the system safe enough for deployment as a whole?

Continuous improvement:
How can data be used to continuously improve the system?

“We collect data to demonstrate that our system is safe across all scenarios,” Lars explains.

Every new software version goes through this cycle again. Testing is therefore not a one-time step, but a continuous process.

The journey of an autonomous vehicle to public roads is built on a wide range of testing environments and methods. Simulations, end-to-end testing, test track evaluations, and public road testing each provide unique insights and together enable a comprehensive assessment of the autonomous system. This highlights an important point: testing is not just a technical step, it is the foundation for trust in autonomous mobility.