Explainable Knowledge-Based AI Could Unlock the Next Era of Autonomous Driving by Transforming Compliance, Transparency and Global Regulatory Approval

Ignite by FORVIA HELLA and Oxford Semantic Technologies are combining explainable AI with machine-readable traffic laws to address one of autonomous driving’s greatest remaining challenges—proving that autonomous vehicles are not only capable of making safe decisions, but also of explaining exactly why they made them.

Executive Summary

The automotive industry has spent decades developing increasingly capable autonomous driving technologies. Today’s advanced driver assistance systems can identify pedestrians, cyclists, road signs, lane markings and hazards with remarkable accuracy, while artificial intelligence continues to improve perception, prediction and vehicle control at an unprecedented pace.

Yet despite billions of dollars invested in autonomous driving, widespread deployment of Level 3 and Level 4 autonomous vehicles has remained frustratingly slow.

The reason is becoming increasingly clear.

Manufacturers are no longer judged solely on whether an autonomous vehicle can drive safely. They must also demonstrate—with robust evidence—that every decision their vehicle makes is logical, traceable and compliant with local traffic laws.

This challenge has become one of the biggest barriers preventing the transition from driver-assisted vehicles to genuinely autonomous mobility.

Ignite by FORVIA HELLA, the software venture dedicated to Software-Defined Vehicles (SDVs), believes the answer lies in knowledge-based AI.

Working with Oxford Semantic Technologies (OST), the Samsung Electronics-owned artificial intelligence company behind the RDFox® reasoning engine, the partners have demonstrated a new software architecture capable of transforming human-written traffic legislation into machine-readable rule sets that continuously verify autonomous vehicle behaviour.

The result is an explainable AI framework capable of providing regulators, engineers and manufacturers with something that traditional machine-learning systems struggle to deliver—deterministic evidence explaining why an autonomous vehicle made every decision.

If successful, the collaboration could represent one of the most significant advances in autonomous vehicle certification since automated driving first entered mainstream automotive development.

Autonomous Driving’s Next Challenge Isn’t Driving

For much of the last decade, autonomous driving has been measured by one simple question:

Can the vehicle safely navigate increasingly complex roads?

Automotive manufacturers and technology companies have answered that challenge by building remarkably capable artificial intelligence systems capable of processing enormous amounts of sensor information every second.

Today’s autonomous driving platforms combine cameras, radar, lidar, ultrasonic sensors, GPS, high-definition maps and powerful onboard computing to create an increasingly accurate understanding of the world around the vehicle.

Deep neural networks continuously analyse this information, recognising objects, predicting movement and determining the safest path forward.

Every year those systems become faster.

Every year they become more accurate.

Yet despite remarkable technical progress, fully autonomous vehicles remain largely confined to carefully controlled pilot programmes.

Why?

Because driving well is only half the problem.

The much harder challenge is proving—with objective evidence—that every decision complies with road traffic legislation.

As manufacturers move beyond Level 2 automation, legal responsibility gradually shifts away from the human driver and towards the vehicle manufacturer.

Once that responsibility changes, regulators naturally ask a different question.

How can manufacturers prove that autonomous systems consistently obey the law?

The answer cannot simply be "because the AI believes it is correct."

Evidence must become measurable, repeatable and independently verifiable.

From Black Boxes to Transparent Intelligence

Artificial intelligence has become extraordinarily capable.

Unfortunately, it has also become increasingly difficult to explain.

Modern machine-learning systems often behave as highly sophisticated statistical engines.

They recognise patterns.

They estimate probabilities.

They generate outputs based upon enormous quantities of training data.

However, understanding precisely why a neural network selected one action instead of another is often extremely difficult.

This lack of transparency has become known as the "black box" problem.

Engineers may know what decision was made.

Understanding exactly how that conclusion was reached is considerably more difficult.

For regulators, insurers and legal authorities, this creates a significant challenge.

When autonomous vehicles eventually become responsible for driving without human supervision, every significant decision may require explanation.

Following a collision, investigators must understand not only what happened but why it happened.

Without explainability, determining accountability becomes extraordinarily difficult.

Knowledge-Based AI Offers a Different Approach

Rather than replacing machine learning, knowledge-based AI introduces another layer of intelligence.

Instead of relying purely upon statistical prediction, it incorporates explicit human knowledge.

Traffic legislation.

Road regulations.

Safety policies.

Operational constraints.

Best engineering practice.

These become structured machine-readable rules that computers can understand and apply logically.

Unlike conventional AI systems that primarily predict outcomes, knowledge-based AI reasons through problems using clearly defined relationships between facts, regulations and expert knowledge.

Every conclusion can therefore be traced.

Every decision becomes explainable.

Every action can be linked directly back to the rule or regulation that produced it.

That distinction could fundamentally change autonomous vehicle certification.

Building a Digital Rulebook for Autonomous Vehicles

Ignite by FORVIA HELLA’s approach begins with an apparently simple idea.

Instead of expecting AI to infer legal behaviour from data alone, why not teach vehicles the actual law?

Traffic legislation has always been written for humans.

Lawyers interpret it.

Police enforce it.

Drivers learn it.

Autonomous vehicles cannot.

Knowledge-based AI changes that by translating legislation into structured digital rules that software can evaluate continuously while driving.

Dr Felix Kortmann, Chief Technology Officer at Ignite by FORVIA HELLA, says the breakthrough came when the company began viewing autonomous driving through the lens of regulation rather than vehicle capability.

Rather than asking whether vehicles could drive successfully, the company asked whether they could prove that every decision complied with the relevant legislation.

That shift fundamentally changed the problem being solved.

Instead of merely producing intelligent behaviour, the software now generates explainable behaviour.

Why Compliance Has Become the Industry’s Biggest Bottleneck

The automotive industry frequently discusses perception accuracy, processing power and sensor performance.

Increasingly, however, the limiting factor has become regulatory confidence.

Manufacturers can demonstrate vehicles successfully completing millions of kilometres in simulation.

They can show excellent performance during testing.

They can provide safety statistics.

Yet regulators increasingly require something beyond performance metrics.

They need evidence explaining how autonomous systems make decisions.

Traditional machine-learning systems are exceptionally good at recognising patterns but considerably weaker at explaining reasoning.

Knowledge-based AI fills precisely that gap.

By combining logical reasoning with machine learning, manufacturers gain both capability and transparency.

This becomes particularly important as vehicles move from Level 2 systems, where drivers remain legally responsible, towards Level 3 and Level 4 automation where liability increasingly transfers to manufacturers.

The Importance of RDFox®

Central to the collaboration is RDFox®, Oxford Semantic Technologies’ high-performance knowledge graph database.

Originally developed by researchers at the University of Oxford and now deployed across multiple industries, RDFox® combines data with logical reasoning.

Unlike conventional databases that simply store information, RDFox® actively evaluates relationships between information while applying explicit knowledge rules.

Within autonomous driving, that capability creates what can best be described as a reasoning engine.

Every observation made by the vehicle can immediately be evaluated against applicable traffic legislation.

Road markings.

Speed limits.

Right-of-way rules.

Temporary restrictions.

Exceptional circumstances.

All become part of an integrated reasoning process rather than isolated data points.

Instead of asking whether the vehicle merely recognised a situation correctly, the system evaluates whether its intended response also satisfies the legal framework governing that situation.

This creates a transparent chain of reasoning capable of supporting engineering validation, regulatory approval and ultimately public confidence.

For an industry preparing to enter the next phase of autonomous mobility, that may prove every bit as important as the artificial intelligence responsible for driving itself.

A Complement to Machine Learning—Not a Replacement

One of the biggest misconceptions surrounding knowledge-based AI is that it competes with machine learning. According to both Ignite by FORVIA HELLA and Oxford Semantic Technologies, the opposite is true.

Machine-learning models remain exceptionally good at interpreting complex environments. They excel at object recognition, predicting the behaviour of other road users, understanding road geometry and making split-second driving decisions.

Knowledge-based AI addresses an entirely different challenge.

It asks whether those decisions are consistent with traffic law, safety policy and regulatory expectations.

Peter Crocker, CEO of Oxford Semantic Technologies, explains that machine learning provides the capability to act, while knowledge-based AI provides the transparency, consistency and accountability required to justify those actions.

Rather than replacing neural networks, RDFox® introduces a reasoning layer that continuously evaluates whether the AI’s proposed behaviour satisfies structured legal and operational rules.

The result is an architecture that combines the strengths of both technologies.

Machine learning supplies intelligence.

Knowledge-based AI supplies explainability.

Together they offer something neither technology can deliver alone.

Understanding the "White Box"

The industry frequently describes modern AI as a "black box" because engineers often know what decision was made without fully understanding why it emerged.

That becomes problematic during validation.

Suppose an autonomous vehicle unexpectedly changes lanes.

Traditional engineering can determine precisely when the manoeuvre occurred.

Understanding the reasoning behind the manoeuvre is often considerably more difficult.

Knowledge-based AI transforms this process.

Every decision becomes linked to explicit rules.

Developers can inspect which legal constraints influenced the decision.

Engineers can determine precisely why one action was selected instead of another.

Unexpected behaviour can be traced directly to the rule involved rather than inferred afterwards.

Equally important, correct behaviour can also be explained.

That provides development teams with significantly more confidence during software validation and continuous improvement.

As software-defined vehicles increasingly receive over-the-air updates throughout their operational lives, this level of traceability becomes increasingly valuable.

Solving a Global Engineering Challenge

Perhaps the most commercially significant aspect of the collaboration is its ability to simplify global software development.

Traffic legislation differs significantly between countries.

Speed limits vary.

Priority rules differ.

Temporary regulations evolve.

Road markings are interpreted differently.

Historically, manufacturers have addressed this by hard-coding region-specific behaviour into autonomous driving software.

Every jurisdiction effectively becomes another engineering programme.

Every legislative update requires further software development.

According to Dr Felix Kortmann, this approach simply does not scale.

Knowledge-based AI allows legislation itself to become structured software knowledge.

Rather than writing millions of lines of bespoke code for individual markets, manufacturers can create machine-readable rule sets that can be updated independently from the underlying driving software.

The same autonomous driving platform can therefore operate globally while adapting its behaviour according to local regulations.

That has enormous implications for development efficiency.

It also significantly reduces the ongoing maintenance burden as legislation evolves throughout a vehicle’s lifetime.

From Engineering Challenge to Regulatory Opportunity

Autonomous vehicle regulation has become one of the defining issues facing governments worldwide.

While the United States has generally adopted a more self-certification-based approach, markets including Europe, Japan, South Korea and China increasingly require manufacturers to demonstrate compliance before deployment.

That places enormous emphasis on evidence.

Regulators need confidence that autonomous vehicles consistently obey traffic laws under an almost limitless variety of operating conditions.

Knowledge-based AI creates precisely that evidence.

Instead of simply observing successful driving behaviour, regulators can examine the reasoning process itself.

Every conclusion becomes auditable.

Every decision becomes explainable.

Every rule application becomes traceable.

For manufacturers seeking approval across multiple jurisdictions, this capability could dramatically shorten certification programmes while increasing confidence among approval authorities.

Public Trust May Depend on Explainability

Regulation represents only one part of the challenge.

Consumers must also trust autonomous vehicles.

Every widely reported incident involving driverless vehicles influences public perception.

When vehicles behave unexpectedly, questions immediately arise.

Why did the vehicle behave that way?

Could the incident have been avoided?

Would another vehicle have reached a different conclusion?

Knowledge-based AI cannot eliminate every accident.

No technology can.

What it can provide is transparency.

Manufacturers can reconstruct the vehicle’s reasoning.

Engineers can determine whether legislation was correctly interpreted.

Developers can improve future software using explicit evidence rather than assumptions.

That transparency benefits regulators.

It benefits insurers.

It benefits legal investigators.

Most importantly, it benefits public confidence.

Auto China 2026 Demonstrates Strong Industry Interest

The partnership between Ignite by FORVIA HELLA and Oxford Semantic Technologies attracted considerable attention when demonstrated publicly at Auto China 2026.

According to Dr Kortmann, the industry’s reaction exceeded expectations.

Interest came not only from major global vehicle manufacturers but also from autonomous driving software developers and specialist technology companies.

Perhaps most revealing was the industry’s response once engineers understood the concept.

Many organisations had accepted duplicated regulatory software development as unavoidable.

Once they saw knowledge-based AI translating legislation into reusable machine-readable rules, many recognised an opportunity to eliminate substantial engineering effort.

Instead of every organisation independently interpreting identical traffic legislation, a shared reasoning framework offers a far more efficient approach.

For an industry under constant pressure to reduce software development complexity, that prospect proved particularly compelling.

The Future Software-Defined Vehicle

Software-defined vehicles continue to redefine automotive engineering.

Vehicle functionality increasingly depends upon software rather than hardware.

Features evolve throughout the vehicle lifecycle.

Safety systems improve through updates.

Autonomous capabilities expand continuously.

Knowledge-based AI aligns naturally with this future.

Traffic regulations change.

Safety policies evolve.

Operational requirements expand.

Rather than rewriting entire software architectures, manufacturers can simply update structured knowledge rules.

That flexibility will become increasingly valuable as governments continue refining autonomous driving legislation.

The result is a vehicle architecture capable of learning, adapting and remaining compliant throughout years of operation.

Why This Partnership Matters

The collaboration between Ignite by FORVIA HELLA and Oxford Semantic Technologies is about far more than introducing another artificial intelligence technology.

It addresses one of the most fundamental questions facing autonomous mobility.

How do manufacturers prove that autonomous vehicles make the right decision for the right reason?

For decades the industry has concentrated on building increasingly capable AI.

The next phase will almost certainly require equally sophisticated explainability.

Knowledge-based AI offers a practical route towards that objective.

By combining statistical learning with explicit logical reasoning, manufacturers gain the ability not only to develop intelligent vehicles but also to demonstrate why those vehicles behave safely, legally and consistently.

That distinction may ultimately define the transition from advanced driver assistance to genuinely autonomous mobility.

As governments introduce increasingly demanding regulatory frameworks and public expectations continue rising, explainability is rapidly becoming as important as capability itself.

The automotive industry has always embraced innovation.

The next breakthrough may not be teaching vehicles how to think faster.

It may be teaching them how to explain their thinking.

Conclusion

Autonomous driving has entered a new chapter.

For years the industry measured success by perception accuracy, computing performance and miles driven without intervention.

Those achievements remain vital, but they are no longer sufficient on their own.

The future of autonomous mobility depends equally on transparency, accountability and trust.

Ignite by FORVIA HELLA and Oxford Semantic Technologies have recognised that challenge and developed an approach that complements modern AI with structured knowledge, logical reasoning and machine-readable legislation.

The result is a framework capable of transforming autonomous vehicle compliance from an engineering afterthought into a core capability.

If knowledge-based AI delivers on its promise, it could significantly accelerate the safe deployment of higher levels of autonomous driving while providing regulators, manufacturers and consumers with something that has often been missing from artificial intelligence: confidence rooted in understanding.

In an era where software increasingly defines the automobile, understanding why a vehicle acts may become every bit as important as the action itself.

Article Summary

The collaboration between Ignite by FORVIA HELLA and Oxford Semantic Technologies represents a significant step towards solving one of autonomous driving’s greatest remaining challenges: proving regulatory compliance. By combining machine learning with knowledge-based AI through RDFox®, the partners have developed an explainable reasoning layer capable of translating traffic legislation into machine-readable rules, tracing every autonomous decision and providing deterministic evidence for engineers and regulators. As Software-Defined Vehicles evolve towards higher levels of autonomy, explainability may become the critical technology that enables faster certification, greater public trust and widespread commercial deployment of autonomous mobility.