Worn lane markings, shoddy roads, and uneven signage standards make it harder for autonomous cars to figure out where they're headed on many American roads. Poor weather and sunlight at low-angles can also make it hard for cars to discern the path ahead.
Ford thinks it has answers. The company says it has been testing an advanced LiDAR system at its Arizona Proving Grounds that allows cars to detect obstacles at night and mitigate some of the problems caused by situations beyond their control. On Monday, Ford announced at the SAE World Congress that it will be the first customer to buy the new Velodyne "Ultra Puck" LiDAR sensors.
That's not necessarily a surprise. Ford and Velodyne have worked on LiDAR together for a decade, and at CES earlier this year, Ford CEO Mark Fields showcased an early prototype of the LiDAR pucks. The ones now being tested are a finished product created specifically for automotive applications. Velodyne says the third-generation pucks are available in limited quantity now, with full production slated for the fourth quarter of 2016.
Sensors send 2.8 million laser pulses per second to more precisely scan their surroundings, and in this latest iteration, those pulses are more narrowly targeted to a vertical field of view of 26 degrees. "That's pretty convenient for where we need them," said Randy Visintainer, director of Ford's autonomous vehicle program, which will triple the size of its self-driving fleet this year to approximately 30 Ford Fusion hybrids. "Our expectation is we're going to get better coverage right where we need it."
"As I rode in the back seat, I was following the car's progression in real time using computer monitoring. Sure enough, it stayed precisely on track along those winding roads." – Wayne Williams, Ford research scientist and engineer
Combined with high-resolution 3D maps, the new Velodyne LiDAR has allowed Ford to test in complete darkness without headlights on its vehicles. (See picture above).
"Inside the car, I could feel it moving, but when I looked out the window, I only saw darkness," said Wayne Williams, Ford research scientist and engineer. "As I rode in the back seat, I was following the car's progression in real time using computer monitoring. Sure enough, it stayed precisely on track along those winding roads."
While it's ideal to use data from cameras, radar, and LiDar together, Ford said its testing showed the latter can function independently if needed. Albeit in early stages, that's potentially a significant step along the way to putting Level 4 autonomous vehicles – those which can handle the entire driving journey without any help from a human – on the road.
Fields has said he expects the auto industry to have autonomous cars that can handle many, though not all, driving scenarios on American roads by 2020. Though the cost of LiDAR is still too high to be considered for a production car, Visintainer said the technology is getting better and costs are trending downward.
"We think for some of the initial applications that we've envisioned, it is on the right trajectory to be practical in terms of affordability," he said. "And with things like the Ultra Puck, we've developed LiDAR more attuned to automotive applications, and there are potential breakthroughs on the horizon that Velodyne and others are working on."
Check out Ford's autonomous testing in the dark on the video below: