Jonathan Bio: Jonathan graduated from his MSc in Smart, Connected and Autonomous Vehicles at University of Warwick, WMG and has since completed a graduate scheme within the intelligent vehicles group at WMG. Now he works as a project engineer within the intelligent vehicles group at WMG focussing on Sensors and Human Factors. Prior to his MSc, Jonathan completed a BSc in Mathematics.
Jonathan’s research interests lie within simulation & modelling of sensors, and he has recently been working on RADAR modelling, looking at divergence and interference. Previous experience involves simulation of LiDAR chip performance, validation of simulated LiDAR sensors and development of a rain model for LiDAR.

Alberto Bio: Alberto Venturi, is a project engineer with experience in developing data-driven models to capture key performance information of diverse sensors and complex energy systems, and link these models to continuous, dynamic optimization methods. He worked in UK and Germany in the automotive and energy sector, contributing to the development of innovative technologies; at the moment he works on developing testing methodology for verification and validation of autonomous driving systems.

Presentation: A change of perception: The challenges of perception sensors for Automated Vehicles

Abstract: The automotive sector is in transition from what we have been used to, namely Internal Combustion Engine (ICE) vehicles driven primarily by a human driver, to electrified vehicles where the vehicle itself takes control of large parts, or all, of the dynamic driving task. Honda have already released an SAE level 3 capable vehicle on the roads of Japan [1], [2]. However, for this transition to be successful, several challenges must be overcome. Reports from across the Atlantic will highlight the success of ‘self-driving’ technology in optimal weather conditions. Waymo have successfully introduced a fully autonomous taxi service in Phoenix, Arizona due to the dry conditions that are witnessed there [3]. However, as we know, the weather in Britain is far from optimal for large parts of the year, as is the case in many other countries across the world. Therefore, these ‘self-driving’ or automated technologies must be capable in the sunny and dry weather of Arizona but also the wet and drizzly conditions found common in places such as those in Great Britain. Furthermore, extreme weather conditions such as fog, haze, and snow provide more problems and once the sun sets for the day, the vehicle’s automated technologies performance must not suffer.

To drive safely under its own control, a vehicle must be able to sense and perceive the environment around it. Hence, all vehicles with automated driving capabilities must be fitted with a suite of sensors capable of accurately and robustly perceiving the external environment. A key enabler for automated driving in adverse weather conditions is RADAR. Automotive RADAR sensors are seen to be essential by numerous stakeholders to achieving the required perception in suboptimal conditions [4]. They are not affected by light levels so will work just as well at night, and they are reflected, refracted or absorbed less by rain than other perception sensors such as camera and LiDAR. These strengths make it a desired sensor to support safe vehicle navigation in these adverse weather conditions. However, RADAR does not come without drawbacks and several challenges need to be overcome for it to provide the benefits we envisage.

Firstly, the output from a RADAR sensor is fairly low resolution meaning it is not great at detecting vulnerable road users such as pedestrians and cyclists. The performance is also often hampered by multipath interference, creating ‘ghost’ targets. This is where a RADAR beam will reflect off multiple objects and return to the sensor with enough power to be registered as a detection. RADAR interference can also represent a problem, particularly for Frequency Modulated Continuous Wave (FMCW) RADAR and will only get worse over time as we see more and more vehicles equipped with RADAR sensors. To provide the redundancy necessary for SAE level 3 vehicles and above, it is expected that each vehicle will be equipped with 5-8 RADAR. There is often no way of determining whether the signals returning to the FMCW RADAR sensor are reflected signals from said RADAR sensor or from another RADAR sensor entirely. The Intelligent Vehicles research group at WMG, University of Warwick are investigating these challenges for RADAR sensors in the future. Our research will inform mitigation strategies and future guidelines to determine how to best use multiple RADARs in the vehicles of the future.