Services
IoT and Digital Engineering
Bringing Together IoT and Digital Engineering to Create Rich Experiences
Highlights
- Ensuring safety is crucial for building trust in self-driving vehicles.
- Safety is a major criterion for a vehicle’s roadworthiness, and it must be ensured throughout its life cycle.
- Stringent adherence to standards such as ISO 26262 for functional safety and ISO 21448 for safety of the intended functionality (SOTIF) can help assure safety in all possible situations in autonomous systems.
- AI technologies including generative AI can help build the foundation for compliance with the new standard SOTIF.
On this page
A driverless future
In the race for autonomous vehicles, ensuring safety is of utmost importance.
Safety is a major criterion for a vehicle’s roadworthiness, and it must be ensured throughout the vehicle’s life cycle. However, safety remains a roadblock for autonomous driving (AD) vehicles. Reports suggest that AD vehicles are more than twice as likely as traditional vehicles to be involved in accidents. But they also have the potential to reduce fatalities caused by human distraction and fatigue.
Most automotive original equipment manufacturers (OEMs) have achieved partial driving automation and are working toward high-level automation. These advancements rely entirely on algorithms and sensors that lack the intuitive capabilities of humans for decision making. Therefore, designing robust algorithms, sensors and actuators, and focusing on adaptability are imperative. And to ensure safety in all potential scenarios, stringent adherence to safety standards is non-negotiable.
Safety first
Stringent adherence to standards such as ISO 26262 and SOTIF can help raise the bar on safety.
Today, functional safety standards are built into vehicle design at the conceptual phase itself. The emphasis is on systems that continue to function and protect passengers in case of a problem—as opposed to systems that fail silently.
ISO 26262 for functional safety is already a standard at auto players. It provides a risk-based approach to the functional safety of all automotive systems. Through hazard analysis, risk assessments, and other methodologies, it ensures that potential hazards due to possible electrical and electronic systems failures are identified and mitigated. However, adhering to this standard will not be enough when it comes to hazards caused by environmental factors or driver misuse.
Consider a situation where the lateral movement control of an AD vehicle stops working due to the advanced driver-assistance systems (ADAS) not receiving camera sensor data following a breakdown of the communication line inside the vehicle. That’s when ISO 26262 design principles will kick in to address the problem. But if the ADAS system fails because of camera sensor limitations due to bad weather conditions, ISO 26262 will not be enough. Principles of the relatively new safety standard, ISO 21448 for safety of the intended functionality (SOTIF), can help here. Unlike ISO 26262 that focuses on preventing failures, this new standard addresses the safety of the intended functionality. It ensures that a vehicle performs safely under all conditions, expected and unexpected, and even in ‘edge cases’ when the vehicle slips from predefined restrictions to unknown territory.
Along with prioritizing ISO 26262, identifying and mitigating edge cases with SOTIF can help assure safety in all possible situations in AD systems. Together, ISO 26262 and SOTIF can help raise the bar on safety, enabling AD vehicles to tackle not just internal system failures, but also external factors such as sudden changes in weather or road conditions.
Vehicle cognition
AI technologies including generative AI can help build contextual intelligence of self-driving vehicles.
Intelligent systems that power AD vehicles derive their understanding of the environment and essential insights for the safe operation of vehicles from the stream of data from multiple sensors. They need a constant flow of data to function.
Additionally, ensuring compliance to standards like SOTIF will require building a database of edge cases or rare scenarios. These include extreme environmental conditions like fog, rain, snowfall, loss of GPS signal in tunnels, or passenger’s inexperience of the human-machine interface or cockpit leading to misuse—all of which compromises safety. Ensuring SOTIF, for instance, for a lateral movement control of AD cars begins with a thorough understanding of the functional and system specifications, and identification of potential hazards that could arise from sensor functional insufficiencies or extreme weather. To mitigate these risks, modifications such as integrating reliable sensors that work in extreme weather are implemented. These modifications are tested and validated against a database of edge cases to ensure acceptance criteria are met.
New technologies such as AI, machine learning (ML), deep learning, and generative AI can play a crucial role in building the database. They can help overcome the challenge of virtual simulations with limited real-time data. They can also aid in training and testing AD neural networks against the scenarios in database. With the knowledge bank thus created, OEMs and part makers can help achieve the target of reducing unknowns, increase the coverage against known hazardous scenarios, and improve the roadworthiness of autonomous vehicles.
Building confidence
Compliance to safety standard is key to building trust in autonomous vehicles.
The journey toward safe and reliable self-driving cars is fraught with challenges. Compliance with safety standards ISO 26262 and SOTIF can significantly reduce the risks of accidents and associated liabilities. Together, these guidelines can provide OEMs an assurance framework, build confidence in autonomous vehicles, and help deliver on the promise of safe and convenient mobility for all.
