It鈥檚 a Ford Mondeo, but not as we know it. This one has about 拢100,000 worth of sensors, cameras on the roof, and a fairly ordinary computer in the boot, enabling it to drive itself. So far it鈥檚 notched up hundreds of miles on public roads in Oxford and the surrounding area. By September, it will be one of a fleet of six autonomous cars driving to London.
These tests are being run by DRIVEN, a consortium of British companies developing autonomous transport in the UK. The narrow streets of Oxford, packed with buses, cyclists and erratic pedestrians, are a far cry from the wide American roads where companies like Google-owned Waymo and Uber have been testing their autonomous tech.
Taking a ride
I had the chance to ride in one of the cars at Culham Science Park, a research campus near Oxford where Oxbotica, which has developed the software that drives the car, is based. I got in the back with Paul Newman, a researcher at Oxford University and co-founder of Oxbotica. A safety driver sat behind the wheel, ready to take over if needed, while another technician in the passenger seat kept an eye on the software.
Advertisement
On the quiet and familiar roads at Culham, Oxbotica鈥檚 control system Selenium has a relatively easy job. The safety driver pushes a button to pass control to the computer, and the handover happens smoothly. The car takes me on a loop around the campus, negotiating traffic lights, roundabouts and zebra crossings with ease.
There is just one moment of drama, when the car suddenly brakes momentarily, for no apparent reason. 鈥淚t鈥檚 seeing a phantom obstacle perhaps,鈥� says Alex, the control system operator. 鈥淭his car has done 350 kilometres in the last week and it鈥檚 done that once or twice.鈥� The team reviews each mistake like this to see what the system is doing wrong. Later, they tell me a new sensor they were testing got confused by the bright sun, and they won鈥檛 be using this sensor when they take the car on a motorway, where unexpectedly hammering on the breaks could have serious consequences.
Selenium relies entirely on cameras to know where it is, rather than using GPS. This means that a human driver first has to drive the route so that the software can create a map for navigating the route independently.
The technology is clever enough not to get confused by differences in the time of day, weather, or other changes in the environment, Newman says. When Culham was blanketed with heavy snow in March, the system was still able to localise itself using maps it had created in the summer.
Pheasants crossing
Out on the public roads, it has had to deal with inattentive pedestrians, unpredictable cyclists and the occasional pheasant. To detect obstacles like this the cars use LIDAR, a system that uses pulses of light to determine the distance to an object.
DRIVEN will have six cars like this on the roads by the end of June. They will have the ability to communicate with each other via 3G, sharing information about road conditions. 鈥淭he single biggest cause of accidents is humans being rubbish at paying attention,鈥� says Newman. Autonomous vehicles will be much more attentive, and can also learn from each other鈥檚 mistakes by sending every car a software update, he says.
In the US, Waymo has notched 5 million self-driven miles on public roads, while Tesla is gathering data from 300,000 vehicles with some autonomous capabilities around the world. Traditional car companies such as Ford, Volkswagen and General Motors are also investing heavily in bringing autonomous cars to market.
Oxbotica鈥檚 algorithms are getting better all the time, but it will take a few more years to ensure that autonomous cars can operate safely in urban areas with lots of pedestrians. This is one of the biggest challenges for the industry, says Srikanth Saripalli at Texas A&M University. 鈥淚 think technologically the biggest hurdles are in human-pedestrian interaction,鈥� he says.
Read more:
UK鈥檚 first public autonomous taxi trial to begin soon
Self-driving Uber death should halt tech鈥檚 race to the bottom
