Google self driving car gets a driver license will soon be on streets
Quick introduction – self-driving cars and laws
There have been a lot of autonomous/driverless cars or vehicles around now for years, dating back to 1939 with GM starting first to several builders from Mercedes-Benz, DARPA Volkswagen, etc… The self-driving cars are used for projects or challenges, say to successfully drive as long as 3000 miles with autonomous accuracy of 97%-99% or to complete a 7 mile race among other of these cars on a tough track at the DARPA Grand Challenge. Even if some of these car builders want them on streets, there wasn’t a law to make that happen in the US.
On March 1 2012, Nevada becomes the first state in the US to pass law to let driverless cars on street of course after getting a license. The first builder of the car to get a license is Google’s modified Toyota Prius. The best engineers from the DARPA challenge were gathered by Sebastian Thrun (project lead, co-creator of Google street view, director of Stanford AI lab and the first DARPA challenge winner) to work on the Google’s self-driving cars project at Google X Labs – a futuristic research lab where things like self-driving cars, augmented reality glasses are being worked on.
Google Self-driving car
Google’s self driving car will be on streets soon taking street photographs to put together for Google map’s street view feature without a driver. A car-bot, AU 001 to start with!
The license plate looks like the one shown below, but the number for Google’s AV (Autonomous Vehicle) is AU 001 as seen from the image above. As per the Director of Nevada’s DMV department, the infinity symbol before number is to denote the car of future. Red color of the plate, shows that these are test licenses and will be green when its made available to the public.
Video cameras, radar sensors, GPS, wheel encoders and laser ranger finder powers the autonomous car to see the road, direction, traffic, signals and obstacles. This is how the car’s on-board visualization (car’s vision) looks like,
The Lidar (LIght Detection And Ranging) sensor, HDL-64E Lidar sensor that is mounted on the roof of the car helps with obstacle detection and navigation. The unit spins uninterruptedly capturing and building a 3D map of the environment with its 64 fixed mounted lasers with a signal output rate of over million points per second. The image above shows the real time processed view of lidar/radar signal, live capturing video and high resolution detailed google maps of the environment terrain.
In addition to Lidar sensor, 4 Radar sensors are used. 2 on the front – left and right and 2 on the rear. Radar senses the objects, how fast its moving, and at what direction its moving. This enhances the Lidar sensor’s obstacle detecting capability.
GPS and Wheel encoder:
GPS gets the vehicle’s location and wheel encoder gets the car movement details. GPS is not accurate and the vehicle has to rely on detailed maps of terrain and with the radar and lidar, the accurate real time location can then be computed.
Safety drivers are there to ensure safety but also as human-less cars aren’t legal yet except Nevada and Florida. Video cameras near rear mirror is to detect traffic lights. But I guess there must be a couple of video cameras near the lights like the Micro-stang car.
It’s not known about the computer that processes all the signals and executes the driving. The car Sebastian built that won the DARPA challenge ran on Intel Pentium M 1.67ghz processor with several versions of Linux running on them. The program was about 100,000 lines of code that processed the signals of lidar, radar, cameras, inertia units, wheel encoders, gps to control the car’s speed, direction and decision making based on machine learning approach such as to detect shadows, to have a driving behavior – say, when in a four way intersection, waiting when pedestrians crossing the road a few meters from the car, to be aggressive when needed etc…all while obeying the road rules.
A Look inside the car
Sebastian Thrun gave a presentation on this technology and how this would reduce road accidents, congestion etc…at Tedtalk last year.