DEARBORN, Mich. — The instructions for the car were simple: drive to the end of the Ford campus and return. There was no more human involvement in the three kilometre drive.
The Fusion, outfitted with radar, lidar (light distance and ranging), infrared sensors and cameras, activated its turn signals and began the drive. Even though this was on Ford property, there was nothing simple about the route. It’s not a “professional driver on a closed course” kind of demonstration. The streets are filled with regular cars driven by regular drivers and there are left turns through traffic, four- and two-way stop signs, traffic signals and pedestrians.
It stopped, by itself, for the first stop sign and waited for a truck on the cross street to pass and pulled out and made a right turn. At the next stop, a four-way, the car activated its turn signal on approach, stopped, waited its turn and then pulled out and turned left. When we arrived at the first pedestrian crosswalk, which was manned by a Ford staffer to ensure there would be a pedestrian, the car saw him approaching the crosswalk and stopped. It waited until the pedestrian was past before proceeding. At other crosswalks, pedestrians who weren’t staged by Ford approached, activated the flashing lights and crossed. The car stopped and waited.
What was most interesting about it all was Ford’s response to a longtime question about autonomous cars: What if the car can’t see lane markings? It’s a problem that continues to plague today’s lane-departure warning systems. Ford’s answer? Ignore them.
“We don’t even use lane markings,” said James McBride, a Ford engineer. Instead, the Fusion uses high-definition mapping information, with resolution down to 10-cm-square pixels, and GPS to place itself in the lane. It uses the radar, lidar, infrared sensors and cameras to detect what the map isn’t telling it. In many cases, it cross-references map data with the other sensors’ information. By doing so, it can detect changes or unexpected obstacles.
It anticipates rolling terrain, scans out 100 metres for obstacles and traffic and recognizes signs and signals.
The map data is complex and must be compiled by Ford using lidar in conjunction with GPS. What it means is the car’s usefulness is limited to areas Ford has already mapped.
Programming the car is much like entering destinations in a navigation system: You tell it where to go, it figures out the rest. Advanced versions could integrate data from traffic-reporting services, such as SiriusXM Traffic, and choose alternate routes based on congestion.
McBride said other tests of the Fusion have involved picking up Ford staff at the headquarters in Dearborn and driving them, autonomously, to the Detroit Metropolitan Airport, about 20 kilometres away, including about 15 km on the busy Interstate 94.
The Fusion prototypes will be put into service by 2018, shuttling Ford staff around its Dearborn facilities. Ford has committed to having a self-driving car on the road by 2021.
McBride said the prototype Fusions could be put into staff shuttle service today, but the company is waiting until 2018 to build more cars and continue to validate the technology.
Bill Ford Jr., Ford’s executive chairman, in an interview prior to the demonstration, said questions remain about autonomous cars, and revolve around not just legal or insurance issues, but ethics. “That’s going to require some very deep and meaningful discussions. For example, whose lives does it save?”
This is a question futurists have asked since the dawn of robotics. If a collision were unavoidable, and the car had to choose between a pedestrian — arguably with less impact for the car’s occupants — and a tree, which does it hit? On what basis does it choose?
One thing is clear: The dawn of driverless vehicles is soon. Sooner than we thought.