As original equipment manufacturers (OEMs) and artificial intelligence startups continue to invest and develop autonomous driving systems, the perception that people have of the technology has remained volatile at best. Though there are two major factions within society, with one voicing support for self-driving technology and the other skeptical about its current relevance, the few accidents that happened last year during self-driving car test runs have led to many people siding with the naysayers.
To commercially deploy self-driving systems, it is critical to not just concentrate on the technology but also to ease the distrust people and governments have about its viability. David Barzilai, the executive chairman of Karamba Security, spoke about how OEMs could address fears by improving security and exhaustively looking at use cases of driving environments to develop or reinforce positive attitudes about the technology. Karamba Security is an Israeli-based company that focuses on securing autonomous technology from hackers via its cybersecurity software.
“If you take an autonomous taxi in Las Vegas, you see a driver sitting by the steering wheel, doing nearly nothing in most cases. This is because when it comes to the last-mile, there’s usually a driver who monitors the steering wheel, just to make driving safe,” said Barzilai. “Obviously, this is not for the long-term, or otherwise it would be missing the point of self-driving. But what we see now is the use of autonomous shuttles for short-range and well-defined routes, with the vehicles driving around at a slow speed of about 20 miles per hour at the maximum.”
However, apart from ensuring the functional safety of these vehicles, it is critical to increase cybersecurity as well, since autonomous vehicles will be prone to hacking, just like connected cars are a popular target at the moment. Barzilai explained that Europe is well ahead of the U.S. in terms of effecting regulations for improving the safety of self-driving vehicles.
In the U.S., the popular stance of regulators is to trail the industry and expect the market incumbents to work together to draft the relevant regulations. This could be counter-intuitive, as companies tend to have misplaced ideals when it concerns safety. Allowing automakers and self-driving technology start-ups to determine the framework behind regulations might lead to ineffective regulations, putting people at risk.
Europe is more cautious, with regulations for self-driving vehicles insisting on functional safety and also auto cybersecurity. Then again, no country has reached a position where it can scrutinize the ethics behind decisions that autonomous vehicles will have to take every instant while on the road.
“For example, a self-driving car is on the road and it suddenly observes a person on a bicycle crossing its path. In a situation where it would be too late for the vehicle to stop, it will be faced with the decision of either veering to the right and run over pedestrians on the sidewalk, or hit the biker,” said Barzilai. “This is a machine-driven decision, and it will always remain a machine-driven decision. This has to do with ethics, and there’s no regulation under development that can address this issue at the moment.”
Even with the technology, some conditions need to be tweaked for a smooth riding experience. For instance, navigating roads in the last-mile is a challenge as vehicles will have to brake more often than when they are on a highway. Surprisingly, even with millions of miles of test runs and thousands of possible scenarios on record, self-driving vehicles still brake abruptly, making the ride a bit uncomfortable.
Barzilai explained that a fully automated vehicle (SAE Level 5) might not be a reality any time soon, but that Level 4 automation (which allows vehicles to autonomously drive across designated geofenced regions and allow human drivers to take control during the last-mile drive within populated spaces) is on the immediate horizon.
To improve auto cybersecurity, Barzilai insisted that it is essential to invite hackers to “attack” vehicles while they are on a test runs, to make sure all the loopholes on the code are fixed before they are manufactured en masse and commercially deployed.
Karamba Security puts this idea to use by virtualizing autonomous vehicles and deploying them globally for hackers to attack the system. This has helped Karamba uncover vulnerabilities which are relayed to OEMs for them to fix them before the vehicles go into production. By exposing problems early on, companies can ensure higher safety standards, and thus allay fears surrounding autonomous driving technology.