Machine Learning and Robotics Group (MLAIR)

About us

The Machine Learning and Robotics Group (MLAIR) is formed by experts in the field of artificial intelligence (AI), machine learning and robotics.

Our research encompasses AI applications in:

  • health
  • theoretical applications of dynamic systems
  • reinforcement learning
  • and developmental robotics.

These applications have been used in business development, health and engineering.

Robotics Lab presentation during Open Days

Research impact

Virtual Reality headset demonstration to control robot head

Machine Learning and Robotics Group addresses essential topics in the application of AI in society and individuals.

Our research provides means to improve a patient's health, but also affects the consequences for the wider community. Additionally, we apply social and ethical considerations to AI and robotics to reduce the bias and unintended consequences of these exciting new developments.

We have shown that our work can greatly improve the well-being of local and global communities in the UK, Europe, and the Americas.

Leadership

Tjeerd V. olde Scheper

Dr Tjeerd Olde Scheper

Reader in Computer Science

View profile

Membership

Staff

Name Role Email
Dr Arantza Aldea Senior Lecturer aaldea@brookes.ac.uk
Professor Nigel Crook Associate Dean: Research and Knowledge Exchange (ADRKE) ncrook@brookes.ac.uk
Dr Clare Martin Principal Lecturer for Student Experience cemartin@brookes.ac.uk
Dr Alex Rast Lecturer in Computing arast@brookes.ac.uk
Dr Matthias Rolf Reader in Computer Science mrolf@brookes.ac.uk

Innovation and patents

A method of controlling a dynamic physical system that exhibits a chaotic behaviour

Patent number WO2013064840A1

The patent is based on the research presented in the paper Biologically Inspired Rate Control of Chaos. It exploits the ability to control complex physical systems using the nonlinear control method to control combustion engines. The patent also shows the ability to control other non-linear and chaotically perturbed systems, such as wind-turbines, and bioreactors. The proof of concept engine has shown the validity of the approach and the patent covers the innovative method that can allow energy efficiency, maintain desired low emission power domains, and even allow fuel neutral engines. 

Olde Scheper, T. V. S. M., & Carnell, A. R. (2013). A method of controlling a dynamic physical system that exhibits a chaotic behaviour. Patent (Awarded 2018). Retrieved from https://patents.google.com/patent/WO2013064840A1

Wind turbines

Robot control with bootstrapping inverse kinematics

Patent number EP2359989A1

This patent covers a specific method of learning robotic inverse kinematics models by means of Goal Babbling. It originates from a PhD project with the support of Honda Research Institute Europe.

M. Rolf, J.J. Steil, M. Gienger. “Robot control with bootstrapping inverse kinematics”, European Patent EP2359989 B1, 02/2011, granted 07/2013. Retrieved from https://patents.google.com/patent/EP2359989A1

Final year students demonstrating autonomous mini-cars
Final year students demonstrating autonomous mini-cars

Research themes and projects

ContrAI

A project with the law firm Moorcrofts LLP is funded by InnovateUK with £284.000. 

The group contributed machine learning and natural language processing solutions for contract law analysis. 

The commercial outcome is ContrAI, a smart contract management suit in which AI helps to identify important clauses in contracts to make their processing better and more cost-effective. 

Moorcraft and Brookes developing AI for contract analysis
Moorcraft and Oxford Brookes University developing AI for contract analysis

Smart Visitor Management and Flow

A Knowledge Transfer Partnership with UNESCO World Heritage site Blenheim Palace funded by InnovateUK with a total worth of £260.000. 

The project makes use of AI and machine learning for smart visitor management and flow prediction, and is done in collaboration with the Oxford Brookes Business School.

Blenheim Palace
Blenheim Palace illuminated and develops smart visitor experience

Goal Babbling

A main research focus of the group is efficient motor learning with Goal Babbling, which was initially described by Rolf, Steil, and Gienger in 2010. 

Recent experiments include:

  • a research student project on learning aimed-throwing skills (MSc by Research, Jamie Pierce, 2020)
  • a collaboration with the Humanoid and Cognitive Robotics group at CTU Prague investigating the learning of tactile body models by means of goal babbling (Gama, Shcherban, Rolf, Hoffman, 2020 & 2021).
Control of throwing robot arm using reinforcement goal babbling learning
Control of throwing robot arm using reinforcement goal babbling learning

Non-linear multi-objective reinforcement learning with M.O.R.E.

The group is pioneering novel methods to make reinforcement learning intuitive and safe by means of multi-objective learning. 

The new MORE method allows for a balanced achievement of multiple objectives (Rolf, 2020) as well as an effective prioritization of needs when necessary (Al-Husaini & Rolf, 2021, PhD project Yusuf Al-Husaini).

Cozmo robots that can learn to meet multiple objectives
Cozmo robots that can learn to meet multiple objectives

Moral Agents and Social Norms

A key research area of the group, and frequent application context of the machine learning methods is autonomous decision making in moral contexts. 

An ongoing PhD project by Rebecca Raper investigates learning architectures for autonomous moral agents. 

A particular focus is the formation of social norms (Matthew Wilson), as well as particular social norms in the context of robotic applications such as proxemics (Vaswani Bhavnani & Rolf, 2020). 

Eddie the human robot head with lifelike motion
Eddie the human robot head with lifelike motion

Biodynamical Research Project

The Biodynamical Research Project within the AI and Robotics research group develops innovative approaches to problems in the dynamic behaviour in Engineering and Medicine. 

These approaches are based on the tried and tested Rate Control of Chaos method that allows nonlinear control of complex dynamic systems.

The Criticality Analysis method shows that a controlled Self-Organised Critical system can be constructed from RCC controlled networks of oscillators. These can then be used to uniquely represent arbitrary data that allows readily classification without training.

Rate Control of Chaos stabilising spatiotemporal chaotic system into periodic orbits using only local information
Rate Control of Chaos stabilising spatiotemporal chaotic system into periodic orbits using only local information
Controlled Self-Organised Criticality applied to machine learning to create nonlinear representation spaces for classification
Controlled Self-Organised Criticality applied to machine learning to create nonlinear representation spaces for classification

Criticality Analysis of Diabetic Gait in Children (CARDIGAN) project

This method has already been applied within the Criticality Analysis of Diabetic Gait in Children (Cardigan) project in collaboration with the children’s Hospital Infantil Federico Gomez, Mexico, to allow gait of children to be used as markers for their clinical progression during their treatment for obesity. Led by Dr Arantza Aldea and funded by the British Council 2019.

Research impact:

Our research in prevention healthcare for diabetes has contributed to:

  • Enhance the diabetes technology research community, through creation of new networks:
    • The South American diabetes research and patient community has been impacted by the successful application for new funding from Newton and the Academy of Medical Sciences for joint projects with Mexico (CARDIGAN, Coordinates, Able-doc).
    • Scientific impact through participation in scientific and industrial conferences and events. All scientific documents are open access through different platforms.
    • Multiple publications
    • Datasets shared with the wider community for future R&D.
    • Clinical confirmation with wider recognition using Impact Acceleration Funds.
  • Educate the public through the communication of the science. A video to explain the research to the general public has been released on YouTube.
  • Strengthen the relations with related industries, by enhancing innovation capacity and the integration of new knowledge, as well as strengthening the competitiveness and growth of companies;
    • Commercially by the industrial partners and through further scientific research and development by the academic partners
  • Impact on society in general
    • Long-term financial gain associated with improved health outcomes for individuals with insulin independent diabetes
    • Societal improvement with gait analysis of general population health can be developed.
CarCardigan Project logodigan
Cardigan Project logo
Presentation of Cardigan preliminary results during public consultation at New College, Oxford University
Presentation of Cardigan preliminary results during public consultation at New College, Oxford University
British Council funded Cardigan project
British Council funded Cardigan project

Avatar Based LEarning for Diabetes Optimal Control (ABLEDOC) 2020

This Innovate UK-funded ABLE DOC project, led by Cognitant Group had OBU and Hospital Universitario San Ignacio, Colombia as partners. The project aim was to improve the self-management of health among people with diabetes in Colombia. This was achieved by conducting a human-centred design study to evaluate the potential of an avatar-based educational programme to improve awareness and understanding of the condition, the effects of treatment, and strategies for effective management of blood-glucose control. The OBU team was responsible for interdisciplinary research to understand whether an AI system could potentially recommend training material by analysing blood glucose patterns.

Research impact:

The 2016 OECD review of Colombia’s health system emphasises the strategy of prevention and treatment of non-communicable diseases such as diabetes. The outputs from the ABLE DOC project could have an enormous social impact since improving the self-management of patients can only lead to better glycaemic control and improvement in quality of life.

The Abledoc team at project launch
The Abledoc team at project launch

Cloud-based Tool for Diabetes Management and Research in Colombia: Initial Investigation (COORDINATES) 2019-2020

This project was funded by the Academy of Medical Sciences and was a partnership between OBU and Universidad Antonio Nariño, Colombia. The aim of the project was to explore the feasibility of a cloud-based platform for diabetes management and research. It focused on the specific needs in the Colombian population and considered the potential for embedding AI within an open source cloud-based platform. The work has led to the development of a bespoke Spanish language platform (CLOUDI) that is currently been evaluated in a large-scale clinical study.

Research impact:

Although cloud-monitoring is already an integral part of clinical management, there is no platform that takes into account the specific needs in Colombia. The platform could be made widely accessible in order to create impact and benefit to Colombian society.

Coordinates research team developing the cloud based diabetes management tool
Coordinates research team developing the cloud based diabetes management tool