Getting there sooner: a new route to self-driving cars
Manufacturers and regulators still face a long journey, but ground-breaking research by XPI Simulation and the University of Warwick could pave the way for the UK government to allow fully autonomous vehicles on our roads.
Connected and autonomous vehicles – or self-driving cars, buses and lorries, in common parlance – have an enormous potential to deliver significant fuel savings, less pollution, safer roads and fewer traffic jams. So what’s holding them back?
An issue of confidence
Technology, although challenging, is not the only barrier to realising the dream of autonomous vehicles. After all, we’ve had unmanned air vehicles in service for some years. Thales’s Watchkeeper, which is used by the British army, can take off and land automatically, gathering intelligence for 16+ hours at a time. Unmanned underwater vehicles and surface boats are also well-advanced.
The equipment used by these unmanned boats and aircraft, which includes RADAR, LIDAR, GPS, altimeters, gyroscopes, secure communications and imaging systems, is also used in driverless cars. It is grist to the mill for companies such as Thales, XPI’s parent company.
Public confidence is a crucial issue. If something goes wrong with a driverless car on the motorway, many people could die. The stakes are as high as they can be.
No government is going to certify autonomous vehicles for everyday road use without a mountain of evidence to show that they are both predictable and safe, even during a highly improbable sequence of adverse events.
Where we are today
Waymo (Google’s spin off), General Motors, Mercedes-Benz, Uber and many others are already conducting on-road autonomous vehicle testing, with thought-provoking results.
Early data from road tests in the US, a fairly benign environment, revealed that the poorest performing car required human intervention every 1.3 miles. The best, a car from Waymo, required an intervention just once every 5,596 miles - about five times better than its nearest rival.
It is clear that, even if safety performance improves by a factor of ten or even a hundred (and it will), much more testing is needed before driverless vehicles can be certified for use on the road. But, with public safety at stake, sending cars around the streets with a driver babysitting is never going to generate sufficient data. Indeed, the five billion mile calculation in the oft-cited RAND study indicates the sheer scale of such a task – not to mention the prodigious cost of carrying out such activity. Fortunately, there may be a better way.
The role of simulation technology
Innovate UK, a government organisation that was set up to help businesses realise the potential of new technologies, recently awarded funding to XPI Simulation and Warwick University to carry out a feasibility study: Certification of Autonomous Vehicles in Synthetic Environments.
The likes of Waymo already carry out an element of autonomous vehicle testing and development in simulation, but if such simulators are going to be used as part of a formal certification process there needs to be a benchmark of quality for such test environments. How can users be sure that the synthetic environment is an adequate representation of the real environment, and that the vehicle under test is going to behave in a similar way as if it were in the real world?
Examples from other industries exist. Full flight simulators (an area of expertise which has long been a specialism of Thales) are used to train and test pilots. These devices are certified by aviation authorities against a set of objective and subjective criteria, which enables users to be comfortable conducting pilot training in simulation.
If successful, the feasibility study will be used to lay the foundations for the development and certification of all types of unmanned vehicles in any situation and every environment. If something goes wrong then the vehicle can be improved, and the tests run again.
The use of simulation for testing is faster, safer, cheaper, more versatile and more repeatable than any other method. XPI and Warwick University believe – and the government and industry hopes – that it will make truly autonomous driverless cars an achievable reality sooner, rather than later.
Find out more about the Future of Mobility Grand Challenge here, and click here to read the Press Release.