AI Simulator Being Created to Test Autonomous Vehicles

The UK government is reportedly granting £2.7 million (247 million rupees) to private organizations to develop a virtual reality simulator that will be the national standard in testing connected and autonomous vehicles (CAVs). Dubbed OmniCAV, the project is led by Latent Logic in Oxford, the University of Warwick’s Warwick Manufacturing Group (WMG), and 10 other organizations – at a total development cost of £3.9 million (358 million rupees).

OmniCAV will make use of artificial intelligence (AI) trained models of vehicles and pedestrians to create a highly accurate way for manufacturers to train and develop the AI that operates CAVs. Big data teams up with advanced automotive technology such as OmniCAV and will also be fed with data from traffic cameras, near-miss analyses, accidents, and highly-detailed scans of real roads. These are the elements necessary for deep learning, an AI training methodology which we’ve previously detailed here on Analytics Training.

The end result will be an extensive, open-access library of VR scenarios that can theoretically put CAVs through every conceivable traffic scenario, thus giving them the ability to prepare for or prevent any untoward traffic incidents. Progress will be determined by how the CAVs perform during simulations, compared to how real vehicles perform in equivalent real-world scenarios.

As Latent Logic CEO Kirsty Lloyd-Jukes explains, “…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.”

Of course, OmniCAV is far from the only self-driving technology that’s being developed in the global market today. Right here in India, there are several startups seeking to create their own smart and autonomous tech in the automotive industry. This includes Pranav Manpuria’s Flux Auto, which Analytics India Magazine reports is focused on developing more affordable models of driverless truck technology. At a comparatively low cost of $3000 to $4500 (Rs. 214,000 to 321,000 rupees), Flux Auto wants to introduce a democratized autonomous vehicle system that doesn’t make use of the expensive Lidar system and instead relies on the company’s own algorithms which are designed to work with cheaper equipment.

Meanwhile, in the US, the entire heavy trucking industry is being outfitted with smart technologies for both safety and efficiency. The telematics page on Verizon Connect describes how the technology has evolved from vehicle-related telecommunications to GPS-powered and AI-enabled fleet and package tracking technologies, including big data analyses in the form of insurance risk assessment. The massive amount of data that’s already been collected by the American trucking industry would no doubt be useful should the US government choose to develop a localized version of the UK’s OmniCAV.

“We need very sophisticated algorithms to mimic the real world, and that is why we use AI. A resemblance to real-life experience is required, instead of a simplistic representation,” explains professor Paul Jennings of intelligent vehicle development at WMG via The Investor. “Our major challenge is to prove that these vehicles are safe and reliable.”