RoadLoc: A development and test framework for ground-truth vehicle localisation.


"_Where am I?_ This is the main starting point for the operation of many of today's self-driving vehicles. From knowing their position, they can go on to assess what else is out there, and what to do next. To do this effectively is challenging , as autonomous vehicles are incredibly complex real-time systems to design and engineer. As with any such system, it is crucial to be able to _debug_ their development process in order to gain insight and understanding of why varying system configurations perform differently.

In engineering terms, obtaining the vehicle position is known as _localization_. A key element of developing technologies in this area, is to be able to define a **ground-truth** reference against which the performance of the autonomous vehicle can be assessed under different experimental test conditions. This might be in response to changing environmental conditions (e.g. weather), different sensors configurations (e.g. LiDAR plus cameras, and where they are looking) , or in relation to external factors (e.g. cyber-attack).

This project sets out an innovative framework - consisting of a hardware sensor and analytics software - that can be used to measure the localisation of the vehicle between where it **thinks** it is, compared to where it **actually** is on the road. The result can be used by the development teams to quickly and effectively measure the performance of different vehicle sensor and how they are configured to operate, as well as understanding how the software is making decisions.

Crucially this framework requires no additional external infrastructure to operate (unlike enhanced GNSS solutions) and can work under real road driving conditions (i.e. at normal speeds, under varying weather conditions, etc.). Furthermore, it is completely independent of the sensor or software component being evaluated - so can be used to objectively verify and optimise performance by the development team from the very early to late stages of self-driving vehicle development. This will result in an extremely valuable tool for enabling Level 4 autonomy and beyond, as well as independently delivering critical assessment of safety during the development and validation process."

Lead Participant

Project Cost

Grant Offer

Machines With Vision Limited, EDINBURGH £448,848 £ 314,193


Jaguar Land Rover Limited, COVENTRY £83,513
Durham University, Durham £230,239 £ 230,239


10 25 50