AI-powered pedestrians will be used for autonomous vehicles test simulations | Robotics
WMG at the University of Warwick have just begun work with a consortium of 11 organisations lead by Latent Logic in Oxford on a £2.7 million UK government funded project to create a highly accurate virtual reality simulator environment, including artificial intelligence (AI) trained models of pedestrians and road users, to test connected and autonomous vehicles (CAVs).
OmniCAV, which was awarded funding as part of a competition run by the Centre for Connected and Autonomous Vehicles (CCAV) and Innovate UK, will be fed by highly detailed scans of real roads, traffic camera data, accident data and near-miss analyses. These inputs will be used to create a high-fidelity model of real-world roads, which will be populated with realistic artificial intelligence (AI) based road users. This model will used to create an extensive open-access library of VR simulator scenarios to test connected and autonomous vehicles.
OmniCAV will lay the foundations for the development of a comprehensive, robust and secure simulator, aimed at providing a certification tool for CAVs that can be used by regulatory and accreditation bodies, insurers and manufacturers to accelerate the safe development of CAVs.
The project will validate the realism of the simulator by comparing its outputs with data measured for the equivalent locations and scenarios in the real world. This will include tests on proving grounds and open roads.
The project will culminate in a CAV being put through the entire end-to-end OmniCAV testing programme, from simulator-only, to controlled environment, to on-road testing.
Through representation on international standard committees, OmniCAV’s results will influence, or lead to the creation of, new international standards to ensure safe deployment and certification of CAVs.
Professor Paul Jennings, Head of Intelligent Vehicles at WMG and WMG’s Principal Investigator on the OmniCAV project adds:
“At WMG we have long believed that simulation will play a key role in testing and certification of CAVs. Our 3xD simulator provides a platform with which to bridge the virtual world and real world and since it’s commissioning in 2016 has played a vital role in our research strategy. Scenario identification and virtual validation are still major challenges for the CAV industry. In OmniCAV, we have a diverse and capable set of partners to advance the current state of knowledge, enabling the safe deployment of CAVs to provide benefits for us all”
Kirsty Lloyd-Jukes, CEO of project lead Latent Logic, said
“OmniCAV’s vision is ‘CAVs for All’: bringing safer, smarter, self-driving mobility to urban and rural areas. But first we need to know that driverless cars really can handle our challenging road conditions, on country lanes as much as crowded city streets. Virtual reality “driving tests” are the only way of doing this, which is why we’ve brought together these 11 leading organisations to build a world-first, AI-based simulation of real Oxfordshire roads to securely and reliably test autonomous car safety.”
The partners in the project are: Latent Logic (lead), WMG (at the University of Warwick), Admiral, Aimsun, Arcadis, Arrival, Ordnance Survey, Oxfordshire County Council, UK Atomic Energy Agency, and XPI Simulation. Thatcham Research is providing advice as a non-funded partner.
Source: University of Warwick