Is LIDAR Necessary For Fully Autonomous Vehicles?

 

Autonomous vehicles use electronic sensors to gather information about the outside world. Those sensors substitute for the eyes and ears of human drivers. The data they collect is fed to high speed computers who turn it into digital maps that are then used to guide the car along its path. The typical assortment of sensors include ultrasound equipment, GPS input, cameras, radar transmitters, and LIDAR.

lidar autonomous driving system
Hyundai autonomous driving sensor system

LIDAR is much like radar except it used pulses of light instead of pulses of radio waves to probe the surrounding environment. According to to Seeking Alpha, it is accurate to about 3/4″ when travelling at highway speeds while cameras are only accurate to about 4″ under similar conditions. If LIDAR is so accurate, doesn’t that mean it is an essential component for all autonomous vehicles?





Elon Musk doesn’t think so. He has said time and time again that LIDAR is not needed to make Tesla automobiles fully capable of driving themselves. Elon is “full of crap,” says Scott Miller, director of autonomous vehicle integration at General Motors. “The level of technology and knowing what it takes to do the mission, to say you can be a full Level 5 with just cameras and radars is not physically possible.”

Waymo, the self driving branch of Google, relies on LIDAR. In fact, a fight over LIDAR technology was the focus of a lawsuit Waymo brought against Uber after it hired former Waymo engineer Anthony Levandowski. Ford and Baidu have invested heavily in Velodyne, a company that manufactures LIDAR equipment. Bit the components of a typical LIDAR system are quite expensive compared to the low cost of digital cameras, ultrasound sensors, and radar receivers. Seeking Alpha says that high cost is needlessly delaying the roll out of autonomous cars — which have the potential of saving tens of thousands of lives each year by reducing the number of fatalities in highway accidents.

New research conducted at several universities around the world finds that a combination of four cameras and an inertial measurement unit is good enough, at least at parking lot speeds. Further tests will be needed to verify performance at highway speeds, but the distance between cars on the road is usually measured in yards, not inches. Traffic aware cruise control systems do a good job of maintaining safe distances between vehicles.

Seeking Alpha says the new study confirms that Tesla has chosen the correct path forward as it pursues its goal of being the first company to offer fully autonomous cars. While the high cost of LIDAR technology will hold other companies back, Tesla will be gathering data from its entire fleet of cars daily. By 2020, Tesla will have collected information from more than 11 billion miles of driving. Other companies will struggle to amass one tenth as much. Even if the cost of LIDAR drops in the future, other companies will lag years behind Tesla at getting their self driving cars on the road.

Score one for Tesla, says Seeking Alpha. But wait. Last month, a tiny startup company called Strobe announced it had created a new “LIDAR on a chip” system that slashes the cost of the technology by 99 percent. It was quickly bought up by Cruise Automation which itself was purchased by General Motors last year.

So, is LIDAR necessary to self driving vehicles or not? There is no question that more information makes for better decision making but LIDAR is often hampered by low visibility situations — smoke, haze, snow, rain, and the like. Tesla has figured out how to use radar to create highly accurate digital maps, allowing it to de-emphasize its reliance on cameras. The presence or absence of LIDAR may prove to more of a marketing issue than one of safety in the future. “The new Belchfire 5000! Now with 20% more LIDAR!!”

But as a new report from the RAND Corporation points out, as long as self driving systems are better at preventing accidents than human drivers, autonomous technology should be embraced sooner rather than later. The battle over LIDAR will be resolved at some point just as the tussle over what charging standards to adopt worldwide will get resolved. With it or without it, expect traffic injuries and fatalities to decline as computers that never get tired, never drive drunk, never have road rage episodes, and never get distracted become more common.





About the Author

I have been a car nut since the days when Rob Walker and Henry N. Manney, III graced the pages of Road & Track. Today, I use my trusty Miata for TSD rallies and occasional track days at Lime Rock and Watkins Glen. If it moves on wheels, I’m interested in it. Please follow me on Google + and Twitter.

  • Ed

    Cars are currently driven by humans using two eyes and inertial sensing to judge distance by a computed inputs:
    1. recognized object size change in field of view. (object subtends a greater arc when moved closer, etc.)
    2. stereoscopic view adds information to 1)
    3. state of acceleration change in driver’s vehicle

    Dozens of other inputs go into that same computer (brain) to assist in safe driving. For example: is the driver in the car about to enter from a side street into my line of drive looking at me? If not, that car will have more of my focus.

  • Matjaž Ciglar

    Excellent article!