The basic technology needed for a car to conduct itself has been available for some years, but it wasn’t until 2004 that formal trials started with DARPA’s Grand Challenge. That required cars to complete a 150mile course through the Mojave Desert. Unfortunately, none of the entrants managed to go further than eight miles.
The Challenge was issued again in 2005, when five vehicles completed the course, with a car from Stanford taking almost seven hours to win the event.
Since then, autonomous vehicles have caught the attention of leading companies, such as Google, as well as capturing the headlines after a series of accidents.
The UK Government is keen for the country to be involved in a market predicted to be worth in excess of £63billion a year by 2035 and recently announced that it was investing a further £38million in collaborative R&D projects targeting next generation AI and control systems for driverless cars.
One of the recipients of funding is the DRIVEN consortium – which has won an £8.6m grant from Innovate UK. Its ambition is to deploy a fleet of fully autonomous vehicles in urban areas and on motorways. At the end of the 30 month project, the partners plan to run a car autonomously from London to Oxford at Level 4 autonomy. This means the car will be able to perform all safety-critical functions and monitoring road conditions for the entire trip. There will be no passengers in the vehicle. According to the consortium, ‘no connected and autonomous vehicle trial of this level of complexity and integration has ever been attempted’.
Leading the project is Oxford-based AI developer Oxbotica. Dr Graeme Smith, the company’s chief executive, said: “No company, group or consortium of autonomy experts has attempted what DRIVEN is planning over the next 30 months. We are seeking to address some of the most fundamental challenges preventing the future commercial deployment of fully autonomous vehicles. I have full confidence in DRIVEN’s world-leading and internationally respected team of specialists to deliver this project.”
While the goal is the London to Oxford trip, DRIVEN believes it will also shake-up the transportation and insurance industries. Key challenges include: communication and data sharing between connected vehicles; connected and autonomous vehicles insurance modelling: risk profiling; and the cybersecurity challenges posed by data sharing.
The consortium work will use of a fleet of six inter-communicating vehicles equipped with Oxbotica’s Selenium software. Said to be ‘agnostic’, Selenium provides any vehicle in which it is applied with an awareness of where it is and what surrounds it. With that knowledge, the project says, the software can then determine how it should move to complete a task.
Professor Paul Newman, head of the Oxford Robotics Institute – a member of the consortium – said: “DRIVEN brings a host of new questions surrounding the way these vehicles will communicate with each other. We’re moving from the singleton autonomous vehicle to fleets of autonomous vehicles – and what’s interesting to us at the Oxford Robotics Institute is what data the vehicles share with one another, when and why.”
The StreetWise consortium has also been awarded £12.8m to develop driverless vehicle technology. Bringing together Cambridge start up FiveAI, the University of Oxford, TRL and Transport for London, the consortium is planning to have autonomous cars on the road by 2019.
Meanwhile, Bosch and NVIDIA are working on an AI system for mass market cars. The system will allow vehicles to be trained on the complexities of driving, operated autonomously and updated over the air with new features and capabilities.
Bosch’s CEO Dr Volkmar Denner said: “Automated driving makes roads safer and artificial intelligence is the key to making that happen. We are making the car smart.”
The system will be based on NVIDIA’s DRIVE PX technology, including the Xavier single-chip processor, which has been designed to support level 4 autonomous driving. Xavier will handle self driving computation tasks, including running deep neural nets to sense surroundings, understanding the 3D environment, localising the car on an HD map, predicting the behaviour and position of other objects, computing a safe path and car dynamics.
