End-to-End Deep Learning Control for Connected Autonomous Vehicles

Lead Research Organisation: University of Surrey
Department Name: Mechanical Engineering Sciences

Abstract

propose an end-to-end reinforced deep learning for autonomous emergency braking system of connected vehicles. The system would use recorded images as well as steering, acceleration, and braking signals from a human driven vehicle for the learning data in order to self-optimise the performance of an autonomous emergency braking system. The main objective of the project is to enhance the safety of the current autonomous vehicles in safety critical situations.

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512217/1 01/10/2017 31/01/2022
1949054 Studentship EP/R512217/1 01/10/2017 25/09/2021 Sampo Juhani Kuutti
 
Description We have developed techniques for autonomous driving using machine learning, and assessed how the safety of these algorithms can be validated through novel methods.
Exploitation Route The testing frameworks developed in this project are fairly general, and could be adapted to gain insights and better understand machine learning tools in other projects and fields. But this research is still on-going (award is still active), and we expect this framework to further extended to better understand how it can most effectively be used.
Sectors Digital/Communication/Information Technologies (including Software),Transport

 
Description Collaboration with Jaguar Land Rover 
Organisation Jaguar Land Rover Automotive PLC
Department Jaguar Land Rover
Country United Kingdom 
Sector Private 
PI Contribution Collaborated with Jaguar Land Rover in investigation into effective safety analysis of machine learning components in autonomous vehicles. Investigated existing techniques and provided internal reports for partners to summarise the current capabilities, future needs, and research challenges in this domain. Developed novel techniques for safety validation of neural network-driven autonomous vehicles.
Collaborator Contribution Jaguar Land Rover participated in discussions on existing and novel safety analysis techniques, and provided an industry perspective on current tools and private sector needs.
Impact Developed novel techniques for safety validation of neural network-driven autonomous vehicles. This resulted in both internal reports, as well as peer reviewed published papers.
Start Year 2018