Introduction

Vision

To create a world leading scientific programme of data driven AI research and innovation that addresses the unique challenges arising from and towards deployment of Robotics and Autonomous Systems (RAS) technology for solving socially relevant problems across domains in a safe and ethical manner.

Aims

The key aim of this strategic challenge of the Turing's AI programme is to develop and support a world leading portfolio of activities that will lie at the intersection of data driven AI and machine learning, specifically targeted to the robotics and autonomous systems (RAS) domain. This will be achieved through:

  1. Building and funding a core research team investigating ‘fundamental’ algorithmic and computational innovations under three key strands, with a research team lead heading each strand
  2. Developing joint industry projects (JIPs) through deep dive engagements with industry for technologies that are at medium to high 'technology readiness levels' – with at least 50-75% of core funding coming from industry
  3. De-risking deployment through proof of concept implementations in several partner ‘living labs’ that integrate hardware and data processing challenges

Programme challenges

The core research agenda will be focused through three strands (that are crucial but are missing or underrepresented elements in the current RAS and machine learning roadmap).

Each of these strands of research will be closely grounded in the RAS context and will have a grand challenge counterpart that it can enable, if successful. However, we will actively look for synergies where some of the results (for example in explainable and verifiable AI) can be applied to domains beyond RAS such as Internet of Things, medical diagnostics, etc.

Scalable algorithms under constraints

Real time inference requirements, computational constraints of embedded, untethered and mobile platforms, hardware limits (torque, joint) for guaranteeing safety, approximate hierarchical inference for graceful performance degradation.

Methods for efficient multi-agent computations

Intention detection and movement prediction, scalable multi-agent adversarial and collaborative policies, multi-modal sensor aggregation for decision making.

Verifiable, robust and explainable decision-making for multi-modal RAS assets

Enabling secure systems – communication, decision making; understanding and predicting failure modes, developing robustness and multiple failure recovery modes, fault and risk inference through probabilistic modelling.

Impact

The programme's engagement approach will aim to create a tangible pipeline that can take strong research and innovation to tangible deployable solutions. The solutions will deliver concrete benefits to society by closely working with various stakeholders – the government, local councils and industry.

The core research programme will identify and develop fundamental AI and machine learning underpinnings that need solving in the RAS domain. This will engage current EPSRC Centres for Doctoral Training (CDT) in Robotics, Data Science and Artificial Intelligence as well as their key scientific leads. Some of the Turing PDRA fellowships in this area are intended to act as a first opportunity for our brightest CDT graduates to play a leadership role.

In addition, we will leverage several world leading sites with substantial RAS assets in terms of cutting edge hardware platforms, to enable stakeholder testing of proof of concept deployments in living labs. We already have several UK wide national robotics hardware and field testing facilities in Edinburgh, Oxford, Bristol, London, Sheffield, and others. 

In conjunction with the living labs at the Bayes’ Centre in Edinburgh (the site of the Turing's robotics hub), we will work in domains ranging from oil and gas, mining, nuclear decommissioning, construction, smart mobility, high value manufacturing (e.g. aircraft), healthcare and smart assisted living space to galvanise industrial engagements through joint industry projects (JIPs) as well as proof-of-concept demonstrations for de-risking in realistic settings.

Finally, engagement with the government and funding agencies (including UKRI, BEIS and learned societies like RAEng, Royal Society, and Royal Society of Edinburgh) will help shape future research funding as well as policy making for enabling, de-risking and deploying RAS technology with the help of innovations in AI and data science.

Organisers

Contact info

[email protected]