Right now self-driving cars are a technical challenge. No amount of sensors and mapping can currently produce a 100 percent reliable self-driving car, but plenty of companies are working on it. When this technology does hit the market, the inevitable question is going to be “how much extra does it cost?”
Waymo, the Alphabet self-driving car division that was recently spun off from Google, is working on getting that cost as low as possible. According to a recent article from Bloomberg, the company has spent the last 12 months working on “scalability,” which has lead to a “90 percent” decrease in the cost of the LIDAR sensor, which is typically the most costly item in a self-driving car solution.
On a self-driving car, the LIDAR sensor is a spinning cylinder that usually sits on the roof of the vehicle. By bouncing a laser off of an object and measuring the time of flight, LIDAR can tell how far away something is. Thanks to the spinning, these sensors can “see” in 360 degrees. Most self-driving car solutions use LIDAR as the major sensor, giving the car a “big picture” view of the world so it can see pedestrians and other vehicles.
The first public Google Self-Driving Car prototype, built on a Toyota Prius, is a good example of how everything works. The biggest component was the Velodyne HDL-64E LIDAR sensor, which cost a whopping $75,000. The LIDAR sensor needed to be mounted up high to see around the vehicle, so Google mounted it on a large riser. This 360 degree sensing was good for a distance view, but not great at detecting up-close objects, thanks to a dead zone around the LIDAR and obstructions from the car body. To fix this, Google augmented the LIDAR input with several black radar boxes stuck to the front and back of the vehicle. These boxes filled in the blanks for close objects.
Google tried roof-mounted GPS sensors, but GPS isn’t accurate enough for self-driving. So, for movement tracking, Google used a “wheel encoder.” An exposed wire ran out of the roof and down the side of the vehicle and connected to a metal stick, which mounted to a spindle on the wheel. This measured wheel revolutions, which was a dead-simple way of measuring lateral movement. In addition to that, there was a video camera mounted in the cabin, and a computer for recording, processing, a cloud communication. At the Driverless Car Summit in 2012, Google disclosed that its self-driving cars used $150,000 in extra equipment.
Cutting down on costs will be a major factor in getting self-driving cars out to the masses. John Krafcik, Waymo’s chief executive officer, told Bloomberg “We’ve made tremendous progress in our software, and we’re focused on making our hardware reliable and scalable. This has been one of the biggest areas of focus on our team for the past 12 months.” Krafcik also told Bloomberg the new sensor package on the Waymo Chrysler Pacifica is “highly effective in rain, fog, and snow,” which have typically been trouble for LIDAR systems thanks to the reflective nature of water in the air.
If we do a bit of math and apply Krafcik’s “90 percent” reduction claim to the $75,000 LIDAR sensor Google’s self-driving car originally used, we end up with a $7500 price tag. While Waymo isn’t using Velodyne sensors anymore, Velodyne has been hard at work cutting down on the cost of LIDAR sensors, too. The company’s newest sensor, the Velodyne LIDAR Puck, is also down to $8000, and Velodyne is working on even cheaper “solid state” LIDAR solutions that don’t offer a 360-degree view. Cutting costs on these LIDAR also cuts down on the capabilities, though. So the major question is, how does Waymo’s sensor resolution compare to Velodyne’s?
The $75,000 Velodyne HDL-64E is $75,000 because it uses 64 lasers and 64 photodiodes (a laser/photodiode pair is a “channel” in LIDAR parlance) to scan the world. This results in 64 “lines” of data output, which you can see in the image above. The $8000 Velodyne Puck only has 16 channels, so while it is cheaper and less complex, you’re also getting a much lower resolution view of the world. Eventually you can cut a LIDAR system down to something cheap enough to fit in sub-$1000 consumer devices like the LIDAR-powered Neato Botvac robotic vacuum, which uses a single-laser system for a 2D view of the world. At what point does the system become too low resolution to be useful for a self-driving car, though?
Waymo seems to favor a higher-detail view of the world, with Krafcik saying “The detail we capture is so high that, not only can we detect pedestrians all around us, but we can tell which direction they’re facing. This is incredibly important, as it helps us more accurately predict where someone will walk next.”
The Bloomberg article doesn’t dive into the technical specs of a Waymo’s LIDAR sensor. Since Waymo is making its own hardware and software, it can find the right balance between cheap-but-not-too-cheap hardware and pair it with software that can make sense of the data.
Listing image by Waymo
http://arstechnica.com/cars/2017/01/googles-waymo-invests-in-lidar-technology-cuts-costs-by-90-percent/ via http://arstechnica.com #CIO, #Technology