UAV-based Scenario Builder and Physical Testing platform for Autonomous Vehicles

Autonomous Vehicles (AVs) offer the promise of a road environment without accidents, much less congestion and a dramatic reduction in car emissions. In order for this vision to become a reality however, AVs must be proven to be safe to operate under any conditions (weather, traffic, other nonautonomous road users). To ensure safety, researchers all over the world overwhelmingly use a two-pronged testing approach, namely simulation and on-road driving. In this paper we propose a third, intermediate, testing environment using fully autonomous Unmanned Aerial Vehicles (UAVs) that realistically (in terms of behaviour) represent the involved actors (AVs, Pedestrians, cyclists, others) in any possible road scenario. Our proposed testing environment successfully addresses the fundamental problems found in existing AV testing testbeds, namely: cost, reproducibility, repeatability, time-consumption, physical world test automation, individual component testing, availability of realistic physical space for testing. Our proposed testbed was successfully demonstrated during the H2020 EU-funded SECREDAS project, using the underlying beXStream framework for autonomous UAVs.