A computer-controlled car that drives itself. Also called an "autonomous vehicle" and "driverless car," self-driving cars date back to the 1939 New York World's Fair when General Motors predicted the development of radio-controlled electric cars. As TVs and modern appliances emerged in the U.S. in the 1950s, more images of self-driving cars debuted. In the 1980s, experiments detecting the lines in the road were performed in the U.S. and Europe, and in 2011, Nevada was the first state to legalize their use.
Some vehicles today feature partial self-driving capabilities such as Tesla's AutoPilot. They self-drive as long as the human driver is attentive (see image below). For driving assist features in late model cars, see automotive safety systems
Accident avoidance is the major incentive for self-driving cars because the computer can respond to dangerous situations a thousand times faster than a human. In addition, people can arrive more relaxed after a long trip. Vehicles can travel closer together and operate more economically when operating in a smooth flow of traffic. The ultimate manifestation is the reduction of vehicles. For example, self-driving taxis could replace a second car, or a family's self-driving car could take everyone to work and pick them up; a much better allocation of resources when you think about the tens of millions of cars that sit idle all day in employee parking lots. Of course, fewer cars has other implications (see computer ethics
If thousands of lives can be saved each year, self-driving cars will be a huge benefit. However, there are situations that are not so straightforward. For example, when temperatures fall below freezing, daily commuters know where the tricky spots are and slow down. Detection systems have yet to meet every challenge perfectly such as potholes full of snow or rain at night. Interpreting hand signals from a policeman or road worker is going to take time to detect properly.
DARPA Grand Challenges
The U.S. Defense Advanced Research Projects Agency jump-started the industry. In 2004, DARPA offered rewards for the winners of a 150-mile self-driving race in California's Mojave Desert. No vehicle completed the course, but 22 out of 23 finished the next race in 2005 with more curves and narrower roads. In 2007, six teams completed a 60-mile run through urban streets.
Google Self-Driving Car Project
Although most automobile companies are in some stage of R&D for self-driving cars, Google undertook its own project in 2009. Seven years later, Google spun off the technology into a new Alphabet division (see Waymo
Taxi Trials Have Begun
In 2016, Uber and nuTonomy began self-driving taxi trials in Pittsburgh and Singapore respectively. Engineers were present in the vehicle to take over when necessary, but drivers did not talk to passengers in order to give them the full driverless experience. See Uber
and autonomous vehicle levels
The Transition to Driverless Cars
Along with the huge technology challenge, state laws are also being challenged and changed. In 2016, California sanctioned the trials of completely autonomous cars (no steering wheel or brakes) in a private business park.
Whether self-driving cars become mainstream in a few years remains to be seen. However, predictions abound that 20% or more of all vehicles wordwide will be driverless by 2040. As that unfolds, the infrastructure is also expected to accommodate this change. For example, road signs, traffic lights and the very roads themselves are expected to communicate with the vehicles.
The Good News
In the meantime, as a result of all the self-driving R&D, accident prevention in regular cars is becoming more advanced, which is a boon to road safety (see automotive safety systems
). See V2X
, self-driving rig
, autonomous vehicle levels
, semiautonomous vehicle
, virtual traffic lights
and automotive systems
The Self-Driving Add-On
Founded in 2013, Cruise Automation made self-driving kits for Audis. In 2016, General Motors acquired the company to turn the Chevy Bolt into an autonomous vehicle. (Image courtesy of Cruise Automation Inc., www.getcruise.com)
What the Car Sees
Tesla Model S Autopilot - 2017
Self-driving computers are processing trillions of operations per second to identify and catalog their surroundings as in this example. See NVIDIA DRIVE
. (Image courtesy of NVIDIA Corporation.)
The Sensing Technologies
The Tesla shows the driver what it detects in the environment. The first self-driving option in a production car, drivers must put their hands on the wheel at least once every 30 seconds. Autopilot alerts the driver when cars are too close, and it automatically changes lanes and parallel parks. See semiautonomous vehicle
Not Self-Driving, But a First Nevertheless
Self-driving cars use a combination of cameras, radar, LIDAR and sound to detect the lanes and objects on the road. This shows their usefulness from worst (red) to best (green). See radar
. (Image courtesy of Phantom Intelligence, www.phantomintelligence.com
Equally revolutionary in 1478, Leonardo da Vinci attempted to build a self-propelled vehicle with coiled springs, but it never worked all that well. This replica is in IBM's conference center in Palisades, New York.