Abstract

The central aim of the EPSRC/Jaguar Land Rover (JLR) TASCC (Towards Autonomy, Smart and Connected Control) programme is to expand research capability to move from enhanced driver assistance towards the central challenge of a fully autonomous car. In the Path-CAD project (Pervasive low-TeraHz and Video Sensing for Car Autonomy and Driver Assistance) the Universities of Birmingham, Heriot-Watt and Edinburgh are working together with JLR on the often-overlooked problem of sensing and algorithmic processing in bad weather, including fog, snow, mist, smoke, rain, night and other unforeseen conditions.

To that end, Andrew will give an overview of the several different research tasks that lead to a successful prototype. This includes the design and manufacture of new radars in the low-THz frequency band, specifically at 150GHz and 300GHz. Allied to hardware development, the talk will address several processing strategies that lead to higher resolution and 3D radar (azimuth, elevation, range), including TDOA (time difference in arrival, synthetic aperture radar, Doppler beam sharpening and compressed analysis).

However, Professor Wallace says “we should not wholly discard optical technology. Therefore, we are also working on sensor fusion, combining optical stereo, LiDAR and radar adaptively in response to the weather. The primary goal is scene mapping. i.e. building a map of the environment around the car. We also wish to place and recognise other traffic participants, cars, bicycles, pedestrians and so on. To that end we have looked at neural networks applied to infrared and radar data, and at how information may be transferred from one network to another modality, so called ‘transfer learning’”.

The project is at the half way stage. In this talk Andrew hopes to show several examples of recent results, and how far they have progressed towards the project goals.

Biography

https://researchportal.hw.ac.uk/en/persons/andrew-michael-wallace