The concept of artificial intelligence may have once been far-fetched, but in 2018, it’s everywhere you look. We invite AI-equipped devices into our homes, into our pockets, and even into our motor vehicles. Fully autonomous vehicles seem to be the way of the future, but self-driving cars haven’t been completely embraced by consumers quite yet.
Although the media has been quick to report on autonomous vehicle mishaps, we don’t hear nearly as much about their successes. There’s been a massive amount of progress made in this arena. And yet, there’s still one challenge self-driving cars have yet to overcome: construction zones.
For human drivers, encountering a construction zone is a surefire source of frustration. For autonomous vehicles, it’s downright confusing and potentially dangerous. Although their algorithms may be powerful, they’re not quite intelligent enough to predict the set-up of a given construction site. In fact, Google and Delphi report that road construction is one of the most common reasons human engineers take control of the wheel during autonomous vehicle testing.
Simply put, construction sites are too unpredictable for current self-driving cars to assess. Even if the technology could accurately discern the physical barriers in a zone, it’s not able to follow directions like a human can. When construction workers have to direct traffic, they rely on hand signals and other non-verbal cues to get the message across to motorists. Autonomous vehicles aren’t yet equipped to read those cues. They may understand roads in the best driving conditions — but they’re currently hopeless at maneuvering around equipment, roadblocks, traffic cones, and other chaotic components we tend to take for granted.
To make matters more complicated, construction zones aren’t always laid out as planned. No two sites are the same, and the same site may change significantly from day to day (or even throughout the same day). What’s worse, the majority of state DOTs won’t update databases that could provide valuable information to self-driving vehicles. Because this data is extremely difficult to access and share, that makes even location tracking a remote possibility.
And that that’s a huge problem for self-driving cars, which rely on data analysis to “learn” and avoid safety risks. Although self-driving car manufacturers would like their vehicles to avoid construction sites altogether, that’s not possible without the cooperation from transportation departments. Without that shared information, autonomous vehicles have to fend for themselves — which could potentially put other drivers and construction workers in harm’s way.
For now, one solution is to have humans involved, one way or another. When a self-driving car happens upon a construction site, it’s going to need a real person to guide it through safely. According to Wired, Nissan planned to set up a remote call center specifically so that humans could help self-driving vehicles in confusing situations such as these. As Nissan’s Silicon Valley research head, Maarten Sierhuis, told Wired: “We will always need the human in the loop.”
Another idea, put forth by the National Highway Safety Administration and the U.S. Department of Transportation, is to give these cars the ability to “talk” to each other and even to infrastructure. With this type of technology, which the NHSA plans to mandata for all new cars by 2020, vehicles would be able to receive information about confusing situations ahead and perhaps how to safely navigate through them. Of course, the success of this idea would require extensive cyber protection and widespread agreements on the essential elements to communicate (and how to communicate them). That makes this possibility a bit more complex than government agencies would have us believe.
In Michigan, a 17-mile-long smart highway is being constructed to test whether IoT communication could help self-driving cars get where they need to go, no matter what. Wearable technology and communicative hardware and signs could send messages to autonomous vehicles so they could navigate along this stretch safely. But constructing a nationwide network of smart highways isn’t likely to be a priority anytime soon.
The most straightforward option, most experts say, would be for self-driving cars to outright avoid construction zones. As we’ve mentioned, that’s not currently possible. But if private construction companies and government agencies were required to document active work zones accurately, self-driving car navigation systems could be easily updated to steer clear of these spots by choosing alternate routes.
The reality is that self-driving vehicles aren’t going to die out; they’re only going to become more prevalent on U.S. roads. Human beings will have to learn to share their world with intelligent technology — and as they do, they’ll need to shift from “the way we’ve always done things” to a method that prioritizes our own safety. The truth is that self-driving cars could make roads safer, but that’s not going to happen if we stand in our own way. By ensuring the tech industry and the construction sector grow together, rather than at odds with each other, the future will be much safer (and our commutes will be more convenient).