Aston University work with RDM Group to develop sophisticated machine vision for autonomous vehicles.

The Company

Aurrigo is a division of the Richmond Design and Marketing Group Ltd (RDM), suppliers of world-class products and engineering services to the automotive industry, including electronic control modules, electrical harnesses, trim and upholstery, and lighting. 

RDM are leaders in the UK’s Low Speed Autonomous Transport System (L-SATS) sector, and introduced the first-ever driverless vehicle in 2017.

These driverless pods are now being produced at RDM in four- and 12-seater versions, along with a Unit Load Device (ULD), which is an automated vehicle suited for use in airport baggage handling or warehouse situations.

The Problem

Self-driving cars is a topic increasingly gathering attention from both the research community and Industry, due to its potential to radically change transport and mobility.

RDM are looking to develop their driver-less vehicles to run at Level 5 autonomy, which means they can run outside confined vehicle free spaces such as airports and closed campuses. This will require the development of an ‘Avoidance Dynamic Path Planning’ capability. To achieve this, the company needs to embed expertise in Machine Vision (MV) and Artificial Intelligence (AI) into its products, which requires complex reasoning and AI-based decision-making techniques that are not readily available outside of academia. 

The Solution

Academics from Aston University’s Computer Science department will support RDM Group to produce a sophisticated detection, recognition and decision-making systems enabling autonomous vehicles to identify and avoid multiple obstacles automatically.

The partnership will allow the company access to Aston University's academic expertise to develop and implement a sophisticated machine vision solution for autonomous vehicles in the L-SATS sector, which will ensure that driverless cars can avoid unexpected obstacles, making their use safer. 

“The expertise of Aston’s academics and their specialist labs are key, allowing us to run detection and recognition of obstacle scenarios. This data will be used in our decision-making algorithms, which will choose the optimal solution based on the environment. This project will allow us to run accurate detection and obstacle recognition scenarios without having to 'field test' our driverless vehicles.” Nick Riddler, Engineering Manager, RDM Group.

This KTP project will adopt a staged approach with the first stage utilising a simulated environment for training and testing. This will enhance the team's understanding, enable the design of suitable model architectures and hyperparameter selection. Once the navigation task using topological maps to represent the 3D world in the simulated environment is achieved, the project will proceed to the second stage, involving on-the-fly visual detection and tracking of obstacles using deep learning and transferring learning from the simulated task to the real-world application.

Further benefits of the KTP include:

  • Delivery of new Low-Speed Autonomous Transport Systems (L-SATS) into mixed environments
  • Higher levels of safety in operating autonomous vehicles
  • Replacement of high-cost detection Systems such as LIDAR with low-cost alternatives, which will improve RDM Group’s competitiveness

Team Aston

The Academic team is led by Dr George Vogiatzis, who has extensive research in the subjects of Machine Vision, Machine Learning and Artificial Intelligence, including deep-learning neural networks and deep-learning methodologies for indexing large video and image collections. While working at Toshiba, Dr Vogiatzis evaluated obstacle avoidance strategies and developed algorithms to improve upon them; expertise that is particularly relevant to this project.

Dr Vogiatzis is joined by Dr Diego Faria who has collaborated on multiple EU and international projects within the fields of Artificial Perception, Cognitive Robotics including Human-Robot Interaction and Autonomous Systems. He is part of the Aston team that has developed significant experience in multi-modal content-based information retrieval
and has applied this technology to commercial applications and investigated deep-learning methodology for indexing large video and image collections.

Both academics have extensive experience of working collaboratively with business to transfer knowledge and create practical solutions. 

Next steps

The KTP is expected to complete in September 2022 – look out for more updates by following us on Twitter and checking our webpage

For more information about Knowledge Transfer Partnerships or an informal chat, email or call 0121 204 4242.