Reservoir Computing Research Sandpit: UK Defence Innovation Funding (Historical Reference)

This page is a historical reference for the UKRI funding opportunity Sandpit: Reservoir Computing for national security and defence. The official UKRI page currently labels the opportunity Closed. Its expression-of-interest deadline was 4:00pm UK time on 16 December 2025, and the page does not announce a replacement round or a new application window. The date remains in the front matter because it is the real closing date for the completed call; historicalReference = true tells readers and the site that this is an archive entry, not a live application listing.

The opportunity was funded by the Engineering and Physical Sciences Research Council (EPSRC) and the Science and Technology Facilities Council (STFC), with the Defence Science and Technology Laboratory (Dstl) as co-funder. The official source describes it as a grant opportunity built around a four-day interactive interdisciplinary sandpit. Applicants were not asked to submit a conventional full research proposal at the first stage. They were asked to complete an expression of interest for the chance to attend the sandpit, work with researchers from different fields and government stakeholders, and develop projects that could then be considered for funding.

What the completed opportunity was for

The central subject was reservoir computing (RC) for national security and defence applications. UKRI describes RC as a non-traditional computing architecture that uses a high-dimensional, nonlinear reservoir to perform computational tasks and implement machine learning. The reservoir’s dynamics can retain information about earlier inputs, making the approach relevant to observed time-series data and systems that need dynamic control. The source also says that RC systems can require less training time and data than more traditional machine-learning architectures, and that reservoirs can be built in physical, electronic, biological or quantum media.

The official description is careful about the limits as well as the promise. RC is not presented as the best fit for every machine-learning problem. UKRI notes that its current high dimensionality and training methods mean it may not match very large, highly optimised deep-learning models. The intended value is in situations where adaptation, low computational burden and control of a changing system matter. The source points in particular to edge networks and small, disconnected devices, including situations where algorithms, architectures and physical substrates might need to be designed together.

The sandpit’s aim was to improve understanding of which combinations of RC algorithms, architectures and substrates work for particular defence tasks, and how RC could fit into a future hybrid computing system. The official page gives an extreme-environment example in which low-power, adaptable computing could support autonomous behaviour in attritable and consumable assets. It also says participants would be introduced to defence and security scenarios by government technology users and encouraged to work across disciplines. The programme was therefore about identifying and testing research directions with stakeholders, not about selling a preselected product or promising a particular deployment.

Funding and project terms

The UKRI page lists a total fund of £1,375,000. It says the full economic cost (FEC) of a sandpit project could be up to £1.37 million and that EPSRC would fund 80% of the FEC. The duration of a successful award was a maximum of 24 months. These figures describe projects emerging from the sandpit; they were not an automatic payment to every person who applied to attend, and attending did not guarantee UKRI or Dstl funding.

The source also records practical support for the event and the resulting projects. Accommodation was to be provided during the residential component. Participants had to make their own travel arrangements, with travel and subsistence costs reimbursed. Where employers could not help, EPSRC said it would cover additional childcare or caring-responsibility costs deemed necessary during the sandpit, in line with UKRI policy. Those provisions were attached to the programme described by the official page and should not be read as a promise that a future or unrelated UKRI call will use the same terms.

If a project was recommended for funding, its leads and co-leads were to be invited to write a full project application describing the activities identified at the sandpit. UKRI said participants involved in projects identified for funding would be informed within two weeks of the sandpit. Funding would remain conditional on a full application and review by the sandpit assessment panel. A funded collaborative project also needed a signed collaboration agreement before its grant could start. The official page expected full applications in May 2026 and final funding decisions in May or June 2026, but this archive page does not claim that any particular project was funded.

Eligibility and security requirements

The opportunity was open to researchers based at a UK research organisation eligible for EPSRC funding. The page refers applicants to EPSRC standard eligibility rules and the UKRI guidance for checking whether an organisation is eligible. The initial expression of interest was for individuals to attend the sandpit, but any resulting award would still follow the relevant institutional and full-economic-cost rules.

UKRI explicitly welcomed people who had not previously worked in the defence or security sector as well as people with existing links. The source lists a broad range of relevant areas, including computer science, artificial intelligence and machine learning, computational biology, computational neuroscience, materials science, physics, robotics, mathematical sciences, signal processing, engineering, high-performance computing, modelling and simulation, digital twinning, psychology, and responsible research and innovation. The purpose of that breadth was to bring together people who would not normally approach the same research problem from the same discipline.

There was an important security condition. Applicants selected to attend had to provide basic personal information directly to Dstl so mandatory security checks could be completed, and attendance was subject to that security screening. This corrects the earlier description on this page, which incorrectly suggested that the initial sandpit did not require clearance. The official wording says screening was a condition of attendance; it does not say that every expression-of-interest applicant had already needed clearance at the time of submission.

The international-collaboration rules were also specific. UKRI said that the UKRI-RCN Money Follows Cooperation Agreement and the UKRI-IIASA agreement did not apply. A submission therefore could not include an IIASA or Norway-based project co-lead as an international collaborator under those arrangements. International collaborators, and UK partners not based at approved organisations, were to be included as project partners instead. These details mattered for any team that later developed a full application.

Sandpit format and scheduled dates

The official schedule listed an opening date of 18 November 2025 at 9:00am UK time and an expression-of-interest closing date of 16 December 2025 at 4:00pm UK time. Selected participants were required to attend the entire sandpit. The in-person part was scheduled for 2 March 2026 to 5 March 2026 in Birmingham, UK, with a virtual sandpit day during the preceding week. UKRI said that the venue and exact format would be provided to successful participants.

The four-day format was designed as a sequence of collaborative activities rather than four days of ordinary proposal writing. UKRI describes the stages as defining the research scope around UK defence and security challenges, building shared terminology across disciplines, exchanging perspectives with stakeholders, working through collaborative idea-generation processes, capturing outputs as research projects, and making a funding decision through real-time peer review at the sandpit. A director, mentors and stakeholders would participate, while the director and mentors would act as independent reviewers and make funding recommendations on emerging projects. They were not eligible to receive the research funding.

Because the event was residential in part and depended on continuous interaction, the attendance requirement was substantive. Someone who could not attend the virtual day and all of the in-person days would not have met the published participation condition. The page also invited enquiries from people with additional needs, including caring responsibilities and disabilities, so that specific provisions could be considered.

What the expression of interest asked

The first-stage application was an EOI survey hosted through the UKRI Engagement Hub. Each question had a word limit displayed in the survey. The published questions focused on the applicant’s approach to open-ended or uncertain research, experience of working across disciplines, communication with people outside their immediate field, current research interests explained for a non-expert, possible contributions to reservoir-computing advances, and what they hoped to gain from and contribute to the sandpit. Applicants were also asked for career-stage and regional information so the panel could consider the diversity of the participant group.

This means the earlier list on this page of a one-page CV, fixed statement lengths, line-manager approval and other supposed required materials should not be treated as official requirements. The UKRI page instead says that selection was based primarily on answers in the EOI survey, alongside consideration of career stages and regions. The listed selection criteria included the ability to develop innovative research approaches, work collaboratively in an interdisciplinary environment, communicate and engage with diverse non-academic stakeholders, demonstrate relevant research expertise and experience, and explain research to non-experts. The panel would also seek a mix of disciplines and backgrounds. UKRI warned that it could not provide individual feedback to unsuccessful applicants.

The survey was not a full technical proposal. UKRI’s general guidance said applications should be self-contained and that hyperlinks should only point directly to reference information; assessors were not required to follow links to conduct their assessment. The page permitted generative AI tools for preparing applications, while cautioning applicants to use them appropriately. Those details describe the closed EOI process and are retained here for researchers studying how the call was structured.

Research themes noted by UKRI

The opportunity did not prescribe one application area. Instead, government stakeholders were expected to introduce scenarios and participants were invited to decide how reservoir computing might address them. The source connected the work to intelligent robotics, artificial intelligence, digital twins, 6G networks, power-efficient data centres and the Internet of Things, especially where fast prediction, computation or dynamic control is required. It also stressed that RC could matter in edge settings and on small devices that are disconnected or subject to power constraints.

The practical research question was therefore one of fit: which RC approach, substrate and algorithm combination is suitable for which defence task, and how can that combination be integrated responsibly? A credible project emerging from the sandpit would need to connect its technical work to a clearly described challenge, explain what would be learned within a maximum 24-month award, and account for capability, resources, ethics, responsible research and innovation, and the needs of the relevant stakeholders. The full-application assessment areas listed by UKRI included fit to the opportunity, project vision, project approach, applicant and team capability, requested resources, and ethical and responsible research considerations.

Status and archive guidance

The official opportunity page remains the authoritative record and is the correct destination for this historical entry. It is closed, and late submissions were not to be considered. There is no current application route described on the page and no next cycle announced there. Researchers looking for a live funding opportunity should not treat the archived deadline as an invitation to apply or assume that the 2026 sandpit schedule has been repeated.

For reference, the official page records the programme’s opening and closing dates, the Birmingham sandpit dates, the maximum award duration, the funding formula and the expected later full-application timetable. It also records the EOI route, eligibility rules, security screening, attendance commitment and assessment questions. Anyone checking whether a new version exists should start from UKRI’s current funding-opportunity search rather than copying this programme’s dates or requirements into a new application.

Official source: UKRI — Sandpit: Reservoir Computing for national security and defence

For questions about the closed opportunity, the source listed [email protected] for specific enquiries. UKRI also listed the Funding Service helpdesk for system or submission questions, but those contacts should not be assumed to accept a new application while the page is closed. The archive is useful for understanding the completed call, its technical focus and its participation model; it is not evidence of an open 2026 or later round.