The primary value proposition for autonomous vehicles — aside from enabling people to realize their greatest "Jetsons"-inspired fantasies — lies in the safety implications of the technology. The operating theory is that by replacing human drivers, autonomous technology will drastically reduce vehicular injuries and fatalities.
But to do that, autonomous vehicles have to drive better than humans in all situations. Four months ago, Ford was the first to begin testing its cars on snow-covered roads and begin training its systems to drive as well as an "expert highly trained driver" would under those conditions. Now, after a series of tests in Arizona, Ford can say its research vehicle can drive better than a human in complete darkness.
Autonomous vehicles rely on three sensors: radars, lidars (a system to detect and measure distance with a laser) and cameras. But cameras, unlike lidars, can’t properly operate in situations with low visibility which includes anything from driving in snow and rain to driving at night. The development of a robust autonomous system that does not require a human to take over (level 4 autonomy), then, depends greatly on building out the capabilities of the other sensors.
"[Testing the car] in complete darkness basically took the camera completely out of the equation," Randy Visintainer, Ford’s director of autonomous vehicle development, told Re/code. "The lidar being the active laser source was able to illuminate the space [in close proximity]. And you can see we could do the localization, object detection and tracking [with just the lidar]. That was the purpose of the test, to show the capability to continue to operate in the absence of the camera."
Ford performed the test on its proving ground in Arizona — where Google will soon be testing its vehicles — and as such the vehicle didn’t have to navigate any other moving objects or people. But it’s still a significant feat for an autonomous vehicle.
During the test, with a person sitting in the driver’s seat but not touching the wheel, the car drove itself around an unlit track with the headlights turned off. The way it works, typically, is that the three sensors gather info and fuse it together to create situational awareness.
But in this case, the system uses a combination of prior maps as well as the lidar, which sends out a signal and gathers data on the range and intensity of objects in its path and around it in order to determine where the vehicle is driving in real time.
Much of the technology — and thus the specs and design of the car — is still in development. But the company expects the initial iteration to be geared toward a commercial ride-hailing service.
"We see the first application of these vehicles as likely in some kind of mobility or ride service," Visintainer said. "We do envision we could be designing a vehicle without a steering wheel once the legal framework, infrastructure and technology is in place. Then you can have an autonomous vehicle that’s operating in a ride service that would not have manual control."
Ford is also looking into developing a vehicle that would allow the owner to switch between a fully autonomous mode and a manual one. But with all that hardware and software development, the company is also working on developing a protocol the vehicle will employ in a situation where all three sensors fail.
"We have a team working on functional safety, which is how to handle different faults," he said. "Will the vehicle be able to get over to the side of the road safely? What level of redundant sensors will we want to have on the road to enable that to happen? We’re looking at those scenarios."
For now, Ford plans to continue to prove out the robustness of its existing technology. The company is announcing that it is the first company to order the most advanced lidar system that Velodyne — the company that manufactures the lidars Ford has been using for the past decade — has just developed called the Ultra Puck. This new line of laser-based sensors has a range of 200 meters and is able to better detect objects that are on the horizon.
With plans to begin testing its autonomous technology in 30 Ford Fusions this year, the company hopes to gather enough data to develop level 4 vehicles that are able to drive better than humans.
This article originally appeared on Recode.net.