Unmanned Systems Technology 020 | Alpha 800 I Additive Manufacturing focus I USVs insight I Pegasus GE70 I GuardBot I AUVSI Xponential 2018 show report I Solar Power focus I CUAV Expo Europe 2018 show report

19 Software for the city The couple of decades between leaving school and the fateful release of his Caterham model into the wild was filled first with studying for a degree in banking and international finance from the Cass Business School at the City University of London, and then with serial IT entrepreneurship, interrupted by a long sabbatical on a yacht. Contracts with many Formula One teams followed, along with regular work for all of the NASCAR manufacturers, the big-budget World Endurance Championship teams and outfits from many other motor sports including Formula E, Super GT, IndyCar and IMSA sportscar racing. Next, the company approached major automotive manufacturers, who were very receptive to the technology because, as Hoyle puts it, the amount of money that can be saved through virtual testing in simulated environments is huge. “You can shave 3, 4 or 5% off the cost of a new platform, and in doing so you can save a quarter of a billion dollars. You don’t have to look for big percentages to make staggering cost savings on those projects.” Hoyle served as technical director at ABS from 1990 to ’97, then as MD of Sigma Computing Associates and, in parallel, SCA Advanced Solutions, which provided similar software, until 1999. At rFpro, Hoyle is the major shareholder but still reports to the board, and he has seven people reporting directly to him who run their own teams. In all, the company has about 35 people, mostly engineers who are divided between software r&d and road modelling. He has an unquenchable passion for the simulation and virtualisation of vehicles. “The fact that we are able to simulate them is fantastic, and the way the industry has been evolving lately is amazing,” he says. He emphasises that it has not just been about the vehicle dynamics. He cites the development of advanced driver assistance systems and autonomy, which brings together AI, large-scale modelling and massively parallel simulation – in this case to prove that autonomous vehicles can operate safely in the real world. “At the moment I find it hard to imagine something more challenging and more interesting,” he says. He notes that the validation of autonomous vehicles is computationally expensive to solve. The real world has to be modelled as accurately as possible, as does the vehicle concerned and its interaction with the world through its sensors and vehicle dynamics. Modelling the world involves integrating Lidar scans, video and photography of real cities and other urban environments, highways and rural areas to create what he terms “insanely accurate” 3D models from petabytes of data, to which is added weather and lighting effects. The rFpro team also includes artists who help ensure that the simulation accurately depicts real materials. “Whenever you look at something you are actually looking at a material, whether it is plaster on the side of a building, bricks, tiles, the bark on a tree or a wooden fence,” he says. “These are all materials that have particular properties that affect the way a scene appears.” Because autonomous vehicles use multiple sensors, it is important to Chris Hoyle | In conversation Unmanned Systems Technology | June/July 2018 As well as major highways and rural roads, rFpro has created the streets of Paris in software. This shows a test of an AI’s driving abilities at a complex junction with vague road markings This video screenshot shows a general view of a hazardous situation, with an inset of the autonomous vehicle’s point of view with false colours applied to the sensor imagery to emphasise hazards