Autonomous vehicle technology is almost ready for widespread deployment—but people aren’t ready for autonomous technology. This is because they don’t yet trust the technology to make decisions fully on its own—thus inhibiting driver-assisted vehicles from transforming to truly autonomous vehicles. We accept a certain level of failures in technology like our laptops, smartphones and Wi-Fi because those limitations are merely inconveniences and we can live with that. Building a vehicle requires safety, security and automotive-quality considerations. But when it comes to technology where our lives are dependent on its performance, we have to hold it to a higher standard. Dependable electronics is a path forward with functional safety as a key feature to achieving this higher standard, and therefore trust. Unless drivers and passengers see autonomous technology as trustworthy, it won’t reach its full potential, as it won’t be used.
If functional safety is one key ingredient in dependable electronics to achieving widespread trust, what can be done to help the industry advance further down this path? Semiconductor companies are working closely with the automotive value chain, including OEMs, tier 1s and software vendors, to bring technology into cars that people will trust as much as they trust that their brakes will work. One way to create this trust with dependable electronics via functional safety is through familiarity—making autonomous vehicles behave more like humans, as opposed to their current decision, making processes which are less intuitive to human behavior. For example, extending a sensor’s range to detect an upcoming red light when a human would see it allows the autonomous vehicle to begin slowing down earlier, thus setting the passengers inside at ease that the car has not only recognized the situation ahead but is also responding to it, rather than accelerating towards the red light when the cars in front of it have stopped.
While driving like a human, the autonomous technology also needs to be unobtrusive—and not disrupt the driving experience in a way that drivers are not used to. Current intrusive technologies like beep-based lane departure warnings are often turned off because they flag too many false positives and drivers get annoyed. Additionally, driving like a human may sometimes mean not fully adhering to laws—because autonomous vehicles are rigorously programmed to follow laws in order to avoid lawsuits and liabilities, they may encounter some issues. On a California freeway, the car is going to come to the end of the onramp and come to a complete stop if it isn’t able to speed or break into a small gap in traffic. It won’t be able to merge into traffic—it will be stuck, dead in the water. Most accidents involving AVs have occurred when the algorithm does something that the human driver behind it does not expect, like stopping too abruptly for the human to react. In order to successfully coexist with human drivers, autonomous technology must be able to drive like a human.
All levels of automation, but especially higher levels like Level 3, 4 and 5, require the driver/passenger’s full trust for adoption. Would the driver be willing to switch their brain off at Level 4 if the system only provides limited functionality in the event of a failure? In the case of Level 3, would they be comfortable re-engaging with the vehicle in a matter of seconds? Furthermore, do fail-operational systems with less dependable semiconductor components really provide the requirement of developing trust with the driver? The answer is no, but this is where dependable electronics become important. In order to fully trust the technology, users need to know the components are functionally safe in addition to being safe in the intended function. In other words, they need to know that the technology will be dependable and reliable, but in a moment of need, the technology can also be trusted to keep its passengers safe.
But what is functional safety in dependable electronics, exactly? There are different levels within this definition. When something is fail safe, it breaks but it does not cause harm. Fail operational means in the event of a failure, the component will still have some functionality for a limited period of time, and high availability means that when something breaks, the rest of the device—the car, for example—can still operate for an extended period of time. As a whole, functional safety is all about how systems react in the event of a failure. This needs to be in place for all classes of vehicles, from economy to luxury, in order to help build user trust in autonomous technology.
Building trust also requires dependable electronics, which serve as the basis for automated systems. Such a system is composed of a control loop that senses, interprets, decides and acts. This control loop needs to be powered and its elements must securely communicate with each other. In order to use such a system in the context of trust and automated driving, its dependability is paramount. Therefore, auto OEMs and tier 1s need to work even more closely with the semiconductor industry to create these dependable electronics that are functionally safe. Key capabilities that OEMs should look for in a semiconductor partner include:
• Agility: Agile suppliers are going to enable new ideas and technologies, as well as enable flexibility further into the OEM’s design cycle.
• Robustness: Early in the design cycle, a high level of innovation and openness to new concepts is ideal. But as the process gets closer to putting an actual vehicle on the road, there need to be some guard rails on that innovation in place, ensuring that the end product that’s deployed into the market is something that’s trustworthy. This is where the heritage of technology companies can help.
As more people experience dependability in autonomous vehicles and those AVs begin to act more like humans, the level of comfort and trust people have in them will increase. It’s currently the unfamiliarity of the technology, along with the fear of the unknown that stands in the way. The foundation of dependable electronics needed to establish this trust begins at the component level. This enables reliable modules for tier 1 customers, and allows for the development of dependable systems and subsystems, which are required to create a safe autonomous vehicle that can interact with its environment. While it will take time to implement more human-like, dependable and functionally safe algorithms and electronics, and time for users to become accustomed to them, the end result will be a safer driving experience for everyone.