Why Testing Automated Driving Is Harder Than It Looks

Most people assume the hardest part of automated driving is building the technology itself. After all, enabling a vehicle to recognize traffic signs, detect pedestrians, stay within its lane, and react to complex traffic situations sounds like an extraordinary engineering challenge.

It is.

But many automotive engineers would argue that the bigger challenge comes afterward. How do you prove the system works safely?

As vehicles become increasingly software-defined, validation has become one of the most important and demanding aspects of ADAS development. ADAS validation is the process of evaluating whether driver assistance functions behave reliably across relevant driving scenarios, environmental conditions, system configurations, and software versions. The challenge is no longer simply creating intelligence. It is demonstrating that intelligence performs reliably across an almost limitless variety of situations.

A Successful Test Drive Isn't Enough

Imagine a new ADAS function successfully navigating a busy urban environment. The vehicle detects traffic signs, maintains safe distances, and responds appropriately to nearby vehicles and pedestrians. Does that mean the system is ready?

Not necessarily. A single test drive represents only one combination of conditions:

• One road
• One traffic pattern
• One weather condition
• One lighting situation
• One vehicle configuration
• One software version

The next drive could look entirely different:

• A child might run into the street.
• Rain might reduce visibility.
• Road markings could be partially faded.
• Vehicles could behave unexpectedly.

The real challenge is not proving a system works once. It is proving it works consistently across countless different situations.

The Scenario Problem

In ADAS validation, a scenario represents a driving situation defined by elements such as road layout, traffic participants, vehicle behavior, weather, lighting, and sensor conditions. Changing just one variable creates a new scenario.

Now multiply those variables across the thousands of roads, weather conditions, traffic situations, and vehicle configurations an automated driving system may encounter. The result is what many engineers call the scenario explosion problem. There are simply too many possible combinations to evaluate using traditional physical testing alone.

The Real Challenge Isn't Creating Scenarios. It's Managing Them

Most organizations no longer struggle to create a handful of test scenarios.

The real challenge begins when those scenarios grow into thousands, hundreds of thousands, or even millions of variations.

Modern ADAS teams must answer questions such as:

• Which scenarios are most safety-critical?
• Which combinations have already been tested?
• How much scenario coverage is enough?
• Which scenarios should be re-executed after a software update?
• How can regression campaigns run automatically as development progresses?

As a result, validation is becoming less about individual test cases and more about managing scenario portfolios, measuring coverage, and continuously evaluating software changes.

This is where simulation moves beyond simple virtual driving and becomes part of a broader validation workflow.

Rare Events Matter Most

Ironically, some of the most important situations to test are also the least likely to occur.

These are often called edge cases:

• A pedestrian suddenly crossing between parked vehicles
• A cyclist approaching from an unexpected angle
• Temporary construction zones
• Unusual traffic behavior
• Poor weather combined with low visibility

These events may occur infrequently in the real world, but they often represent the greatest risks to safety. Finding enough instances of these situations through road testing alone can take enormous amounts of time and effort. In many cases, engineers need a way to recreate these situations repeatedly and under controlled conditions.

Why Physical Testing Alone Cannot Scale

Road testing remains a critical part of vehicle development. Nothing can completely replace evaluating a vehicle in the real world. However, as ADAS functionality grows, the limitations of physical testing become increasingly apparent:

• Vehicles need to be driven extensively
• Tests are difficult to repeat exactly
• Environmental conditions constantly change
• Rare events are difficult to encounter on demand
• Costs increase significantly with scale

Connecting Virtual and Physical Validation

The limitations of physical testing do not mean that road and proving-ground testing become less important. Instead, the opportunity lies in connecting them more closely with simulation.

One example comes from Daimler Truck, which has publicly described a validation approach that combines simulation, proving-ground testing, and public-road evaluation for advanced safety assistance systems. Read the Daimler Truck press release.

In this type of workflow, simulation can be used to explore a broader range of situations and identify relevant corner cases before selected scenarios move into controlled physical testing. The physical results can then provide another source of evidence for understanding system behavior.

This reflects an important shift in ADAS validation: virtual and physical testing are not competing approaches. Their value increases when they are connected within the same validation process.

The Value of Repeatability

Connecting virtual and physical validation also highlights another critical requirement: repeatability. Engineers need more than successful results. They need confidence that results can be reproduced. Consider a near-collision event encountered during testing. If the exact same situation can be recreated repeatedly, engineers can:

• Investigate root causes
• Compare alternative software versions
• Validate fixes
• Measure improvements
• Evaluate sensor performance

Repeatability transforms testing from observation into engineering. Without it, understanding system behavior becomes significantly more difficult.

Why Simulation Is Becoming Essential

This is where simulation begins to change the equation. Instead of waiting for rare situations to occur on public roads, engineers can create those situations virtually. A scenario can be repeated hundreds or thousands of times while controlling individual variables:

• Weather
• Traffic density
• Vehicle speed
• Sensor configurations
• Road infrastructure

The emerging validation model is therefore not “virtual or physical.” It is a connected loop in which simulation helps identify what should be investigated further, while controlled physical testing provides additional evidence for evaluating system behavior.

Simulation also enables a level of automation that is impossible to achieve through physical testing alone. Modern development teams increasingly integrate virtual validation into continuous development workflows, allowing software updates to trigger automated regression campaigns against predefined scenario libraries. This helps organizations identify issues earlier, evaluate changes more consistently, and accelerate development without sacrificing confidence.

The Road Ahead

As vehicles continue to evolve, validation is becoming a data problem as much as an engineering problem. Future development workflows will increasingly combine:

• Scenario-based validation
• ASAM OpenSCENARIO and ASAM OpenDRIVE assets
• Closed-loop simulation
• Sensor and ECU integration
• Continuous regression testing
• Cloud-scale execution

The goal is not simply to run more tests. It is to create validation workflows that continuously measure coverage, identify gaps, and deliver evidence that software behaves safely as systems evolve.

In this future, success will depend not only on building intelligent vehicles, but also on building validation processes that can scale alongside them.

Key Takeaway

ADAS validation is not about demonstrating that a system works in one situation. It is about building confidence across a broad range of relevant driving scenarios, including rare edge cases that are difficult to reproduce through road testing alone.

Explore a Complete ADAS Validation Workflow

Modern ADAS validation requires more than scenario creation. Engineering teams need to manage scenario libraries, integrate existing toolchains, connect sensor and vehicle models, automate regression testing, and scale validation across increasingly complex software releases.

Discover how Cadence VTD supports end-to-end ADAS and autonomous driving validation workflows, from scenario development and sensor simulation to large-scale virtual testing and continuous validation.