Humans and AI could work together to prevent cars from… coming together

Somewhere, right now, there is a vehicle designer working on creating a level 4 or 5 autonomous car. In fact, designers, developers, and technologists all over the world are working right now, sometimes collaboratively, sometimes competitively, to bring self-driving cars to the roadway.

All of those designers face a hugely daunting challenge. They are tasked to create an autonomous vehicle that can successfully navigate all the road and traffic scenarios that their team can imagine. At the same time, they have to create a smart system that is flexible enough to address the scenarios that they can’t predict.

Designing solutions for scenarios you can’t imagine is a bit difficult. That’s why system designers often build in safety protocols that fall back on human intervention. In the industry of autonomous vehicles, this is called a Human-in-the-loop (HuIL) control system.

This style of human-driven backup has been in use for longer than you might consider. Many Unmanned Aerial Vehicles (UAVs) can take off, fly for more than a day, and land again all on their own. While those vehicles busily operate themselves, they often have a crew of pilots on the ground at all times, ready to take over by remote control, should they be needed.

As self-driving vehicles hit the roadway, the need for steering inputs will steadily decrease. When vehicles reach level 5 autonomy, no steering inputs are needed at all. So then, what happens when the vehicle encounters a roadway challenge that the AI cannot safely navigate?

Enter the roadway’s human in the loop

When we call AAA, or when we press the OnStar button, our self-driving vehicles may have a system to call for help. When they make that call, that’s where there’s a potential for a human in the loop. However, this human isn’t in a traditional call center, nor are they at a local mechanic shop waiting to come to help fix a flat tire (though both will surely still exist).

If the industry at large adopts this approach, this human could be sitting in what looks like an elaborate driving simulator. They will have screens on all sides, a stereo headset and microphone, a complex electronic steering wheel and pedals, and likely what looks to be a laptop computer.

When our vehicles dial in, all the cameras and sensors will begin routing information to our autonomous support representative. In their remote simulator, they’ll see everything that the car can see. Once given permission, they’ll be able to take control of the vehicle remotely, navigate through the difficulty, and return the car to its normal function — likely after wishing us a nice day.

While this may sound like we’re getting ahead of ourselves, this type of system is already in the works. Ford is rapidly developing remote-control systems for their future vehicles. Nissan is developing an in-house remote operation center and putting a hefty amount of research toward the requirements and needs of consumers and remote operators.

Meanwhile, California startup Phantom Auto is offering an independent solution, with the aim of creating a high-end consumer experience, while being available to any car on the market.

It’s easy to assume, further in the future, that as the ubiquity of autonomous technology goes up, the need for humans to be licensed drivers will go down. This could mean fewer qualified vehicle operators on the road, leaving a wide opening for remote vehicle operation as a cottage industry.

In the future, if your autonomous car needs help, what will you want it to do?