maybe this was answered, but my only question is how would the car be able to react to the unexpected? Like a cat or kid running out in the road?
Depending on the level of sophistication, a combination of onboard passive and active multi-spectral sensors such as RADAR and LIDAR (a "laser radar" rangefinder/mapper), 360
cameras with nightvision or infrared capabilities (NV or IR not needed with the LIDAR, though), real time satellite navigation (GPS or other), satellite/cellular/radio/WiFi real time updating of traffic/weather/road conditions (like construction) or changes to driving laws/maps, a highly detailed and constantly updated onboard database of physical and street maps & a business/park/location directory, etc.
A lot of vehicles on the road already have many of these features meant to assist the driver rather than take over for them completely (except for extremely brief periods or in emergencies) such as adaptive cruise control, automatic braking, active parking assist or automated parallel parking systems, lane departure warning systems, satellite navigation, front and rear mounted RADAR, nightvision heads-up-displays, front/rear/blindspot cameras and monitors, etc., so we've gradually been making the transition toward more automated vehicles already. In some cases --like with cruise control and ABS-- for decades now.
(pic from first article below)
The Google Self Driving Car operates on a software called Google Chauffeur which can operate in most modern computerized cars with some tweaks, and unless already factory installed in the car, the sensor/guidance systems (LIDAR 3D mapper/rangefinder, GPS, cameras, WiFi, cellular, onboard computer) come in a rooftop rack that clamps on and can be retrofitted to any car (so far Google has tested it on Toyota Prius hybrid hatchbacks, Audi TT sports cars, and Lexus RX450h hybrid SUVs). You could turn your normal car into a driverless car with a visit to the mechanic for half a day in the near future conceivably, albeit at great cost initially before the tech becomes more commonplace and cheaper to manufacture.
Google's driverless car has now driven more than 500,000 miles in the US without causing an accident – about twice as far as the average American driver goes before crashing.
http://www.herald.ie/news/public-to...-cars-is-the-right-road-to-take-29886052.html
So far the dozen or so Google SDCs have only been involved in two accidents, one where it rear-ended another Prius in a minor fender-bender near Google HQ while the human driver was in control of the vehicle, and one where it was rear-ended by another vehicle while stopped at an intersection, so neither time due to the automated driving system.
Google SDCs have driven autonomously down San Francisco's infamous snake-like
Lombard Street, through heavy traffic and fog on the Golden Gate Bridge, in the winding mountain roads and in icy/snowy conditions around Lake Tahoe, and through vast stretches of California/Nevada desert in the heat without a problem, among many other locations and environments they've driven in.
In answer to the earlier question about what laws have been passed regarding autonomous cars so far, Google SDCs have so far been declared street legal for fully autonomous road driving (albeit just for testing purposes so far, not for public sale yet) in California, Florida, and Nevada; and in Michigan with a mandatory driver. Texas is considering legislation allowing driverless vehicles and is addressing considerations about who would be liable in case of an accident: the driver or the company who manufactured the automated driving system. So far, for liability purposes until the concept is fully proven and trusted, Google has chosen to keep a human engineer and driver present while driving on the open road. They've just recently reduced that to only a driver, and will soon do it fully autonomously, as the law in three states already allows. The car frequently drives itself without a manned presence on the Google Campus, however.
Google's SDCs are not by any means the only driverless cars out there, they are just the most successful (legally and numerically), well-traveled, and longest operating so far, which is why I chose to focus on them.
Here's some more info if you're interested:
http://en.wikipedia.org/wiki/Google_driverless_car
To get back to
FSM's question above, here's how the Google driverless car has already reacted to a deer in a forest at night before the human "driver" (who was not in control at the time) was even aware of the deer's presence:
Some of this technology is already used in high-end cars such as the Mercedes S500, which uses onboard radar and 3D stereoscopic cameras to gauge the distance from other cars.
The position, direction and speed of a car on the road can be measured using satellite navigation systems that are already in widespread use.
Google is testing a "self-driving car" in the USA which contains cameras, radar, antennas, and satellite navigation technology. It has a human driver who can intervene in the event of an emergency. It has carried out demonstrations showing the car bringing a man with 90pc sight loss to buy a taco and pick up his dry cleaning.
The 'New Yorker' magazine recently reported how the 'Google Car' never gets drowsy or distracted and never wonders who has the right-of-way on the road. On one occasion, it slowed to a crawl while travelling through a wooded area at night. This puzzled the human driver on board until he spotted a deer walking along the side of the road -- which the car had spotted in advance using its "see-in-the-dark" systems.
http://www.independent.ie/business/technology/how-they-work-driverless-cars-29884532.html
Finally, here are some pros and cons of driverless cars, with there being significantly more and better pros than the few cons, most of which revolve around human concerns rather than tech concerns anyway. Although, the risk from hacking, privilege escalation (jailbreaking) that can result in illegal and/or unsafe modifications, software bugs, satellite/communications interruption, viruses and spyware, and other technical glitches is a potentially very real one that would require strong security measures and signal reliability, and extensively beta-tested upgrades. You wouldn't want some buggy Microsoft-style "upgrade" to your car that takes a couple of service packs before it starts working smoothly. That kind of thing can't be rushed and has to be taken care of well before it reaches the general public's cars themselves, which unfortunately (but necessarily) limits your ability to react swiftly to new security threats.
Potential advantages
An increase in the use of autonomous cars would make possible such benefits as:
-- Fewer traffic collisions, due to an autonomous system's increased reliability and faster reaction time compared to human drivers.
-- Increased roadway capacity and reduced traffic congestion (due to reduced need for safety gaps), and the ability to better manage traffic flow.
-- Relief of vehicle occupants from driving and navigation chores.
-- Higher speed limit for autonomous cars.
-- Removal of constraints on occupants' state – in an autonomous car, it would not matter if the occupants were under age, over age, blind, distracted, intoxicated, or otherwise impaired.
-- Alleviation of parking scarcity, as cars could drop off passengers, park far away where space is not scarce, and return as needed to pick up passengers.
-- Elimination of redundant passengers – humans are not required to take the car anywhere, as the robotic car can drive independently to wherever it is required, such as to pick up passengers or to go in for maintenance. This would be especially relevant to trucks, taxis and car-sharing services.
-- Reduction of space required for vehicle parking.
-- Reduction in the need for traffic police and vehicle insurance.
-- Reduction of physical road signage – autonomous cars could receive necessary communication electronically (although physical signs may still be required for any human drivers).
-- Smoother ride.
Potential Barriers
In spite of the various benefits to increased vehicle automation, some foreseeable challenges persist:
-- Technology Risks, such as software reliability and cyber security.
-- Implementation of a legal framework for self-driving cars and establishment of government rules and regulations.
-- Liability for damage.
-- Resistance for individuals to forfeit control of their cars.
-- A car's computer could potentially be compromised, as could a communication system between cars.
-- Reliance on autonomous drive produces less experienced drivers for when manual drive is needed.
-- Loss of driver-related jobs.
http://en.wikipedia.org/w/index.php?title=Autonomous_car&printable=yes
