Americans were astonished to see a fleet of robot cars driven entirely by computer software around the San Francisco traffic system this month. While it sounds like something out of science fiction, industry experts believe cars may be driven without human intervention far sooner than we imagine.
Google converted seven Toyota Prius models, adding a funnel-like cylinder on to their roofs, then sent them to cover 225,000km of California's public roads. But although the cars were effectively driven by computer software, a human driver was present in each to override the system in case of an emergency. "We've taken the safety of the public, our drivers and our equipment with the utmost seriousness since the start of the project. Every car has two people inside: a specially trained safety driver who monitors road conditions and traffic, and a specially trained software operator who monitors the computer system," says Oliver Rickman, a Google spokesman.
Google says that by using computers to reduce human error on the roads, it will be possible to save lives and make motoring far more efficient. "This is about making cars safer and more efficient. Automation could: make cars safer - more than 1.2 million people are killed every year in traffic; reduce the nation's energy consumption by several percentage points; free up substantial time every day, 52 minutes per working American; triple the capacity of the highway system; and enable entirely new models of car sharing," says Mr Rickman.
Analysts believe the new breed of autonomous cars could also bring global benefits in the form of reduced fuel consumption and road traffic circulation. "General Motors speculates that by 2018 there will be robot cars on sale," says Andrew Jackson, an analyst at Datamonitor, an international research company. "General Motors, BMW and Volkswagen are all testing robot cars. The difference is that, unlike Google, they are doing it in controlled conditions and not on busy roads."
He says the widespread adoption of autonomous vehicles, also called "robot cars", could have repercussions for the global consumption of oil. "Fuel consumption could be radically affected if cars are run at pre-determined lower speeds. The operation of vehicles at their most efficient speeds would cut fuel consumption, and cars driving closer together without erratic behaviour could reduce 'phantom bottlenecks' caused by driver error," says Mr Jackson.
Ford also thinks there is a future for robot cars but warns there are several stages of technological development that need to be completed. "We believe there is a huge opportunity to work with Google to build future safety features," says Wes Sherwood, a safety communications manager at Ford. "We are working with other auto makers and with governments who are all promoting vehicle-to-vehicle communications in the next five to 10 years. Using Global Positioning Systems (GPS) and short radius wireless networks (Wi-Fi) will enable cars to communicate with one another just as people talk to one another over mobile phones."
Ford has already built a test facility in Dearborn, Michigan, in the form of a "smart intersection" that enables cars to communicate with traffic lights as well as with other cars. But Ford is adamant that vehicle-to-vehicle communications must be thoroughly tested and widely deployed before the industry can consider manufacturing truly autonomous vehicles. Car makers such as Ford are interested in partnering with Google because of its global mapping technology, which is a crucial element in robot cars.
"Although this seems a little out of left field, mapping is a crucial part of autonomous driving as, ironically, before you can programme an autonomous car to navigate a highway, the vehicle must have already made the journey while being manually driven so it can remember the route for future reference," says Mr Jackson. Contrary to reports that Google's robot cars are powered by Android, the company's mobile phone operating system, Google has actually combined its mapping with other technologies originally created for US defence projects to develop its autonomous cars.
"We have developed technology for cars that can drive themselves," says Sebastian Thrun, a Google software engineer. "Our automated cars use video cameras, radar sensors and a laser range finder to 'see' other traffic, as well as detailed maps [which we collect using manually driven vehicles] to navigate the road ahead. This is all made possible by Google's data centres, which can process the enormous amounts of information gathered by our cars when mapping their terrain.
"To develop this technology, we gathered some of the very best engineers from the Defence Advanced Research Projects Challenges, a series of autonomous vehicle races organised by the US government." Analysts also believe Google is taking advantage of a whole range of advances in car manufacturing in recent years. "Car controls are progressively moving towards control-by-wire technologies, with the steering being one of the only true mechanically controlled components remaining in cars these days," says Mr Jackson. "Throttle-by-wire has the pedal movement being monitored by onboard computers which relay the information to the engine management system which controls the engine. These digital controls could just as well be directed by a computer without the need for the pedal to be depressed."
But the jury is still out on just how far drivers will be prepared to trust in-car computer systems with their lives on busy roads where drivers do not always behave rationally. "Google's safety argument boils down to reducing human error. The big question is whether a software-based system can account for that," says Mr Jackson. "Google already admits that, on occasion, its drivers have had to override the computer system in cases such as a cyclist unexpectedly jumping a red light.
"The inherent irony of this technology is how can the software running the vehicles be free of human error when humans write the software?"