The idea

In April 2022 we were asked to start shaping our ideas for the the next round of CDTs. In this post, we will expand on our idea for a CDT in Geospatial AI, why we think it is needed, and why we believe we are the right team to deliver it. We’ve presented this through answers to a series of questions below:-

1. What is the proposed Theme/Topic for your CDT?
   Geospatial data considers location-based information on objects, environments, events, and their trends. Applications span energy, agriculture, healthcare, transport, etc. Data-types are multiple, capturing diverse aspects at different resolutions and times, presenting huge interpretation challenges.
   Our proposed CDT considers how AI can improve geospatial data and analysis in-line with non-academic partner needs, and how geospatial data can help answer fundamental AI questions. Cohorts will investigate the pipeline supporting intelligent policy and decisions, from data acquisition to generating actionable information by exploiting AI. Projects and training will be co-developed with partners to nurture the innovation, entrepreneurship, principles, and diverse perspectives needed to confront societal challenges of sustainability and inequality.

2. Why does the UK need 50+ Doctoral Students trained in the area you are proposing?
   The global Geospatial market is expected to hit >£100billion by 2028 with 12% growth/year. The Geospatial Commission’s UK Geospatial Strategy (2020-25) [1] estimates a value to UK economy of >£11billion/year through improved construction efficiency and prevention of damage to underground/hidden infrastructure (land/offshore), improved transport (land/sea/air) and achieving net zero targets. Location data and analytics has already disrupted industries, with ‘taxi hailing firms, delivery services and location-based search engines transforming cities and connecting communities’ [1]. The strategy identifies sensors and AI as key technologies underpinning these opportunities, recognising additional far reaching environmental and social benefits. The Chancellor’s 2022 spring statement [2] also announced £117 million investment in 1000 AI PhDs.
   This CDT helps fulfil UK government ambition by training cohorts of AI specialists with skills in acquiring, analysing, communicating, and making decisions from geospatial data. The diversity of sectors benefiting from geospatial data provides the capacity for innovative thought leaders to be embedded in many organisations (e.g., UK Geospatial Strategy call for evidence had >200 responses). These need trained individuals able to adapt AI and geospatial skills to different domains to manage ever-increasing data volumes and develop new products, services, and sources of information to support growth.

3. What skills will a Doctoral Student coming out of your proposed CDT have which they would not otherwise get from undertaking a traditional DTP PhD?
   Students will become experts in Artificial Intelligence/Machine Learning and Geospatial science. Added value comes from engagement with peer projects through cohort seminars and joint training activities. These develop a breadth of knowledge around individual expertise, with students understanding motivations and constraints that apply to use cases beyond their own – a key skill for future leaders who can adapt to challenges throughout their careers.
   Recruitment will target cohorts with diverse backgrounds (computer science/engineering/environmental science/mathematics/archaeology) and cater for part-time enrolment (e.g., 50% working at partner industries like the Mobility DTP). Cohorts will be initially co-located, taking specialised modules at Masters level (e.g., machine learning, geospatial data and analytics, data capture, regulation, and ethics), adapting options to match their backgrounds. These will be opened to industry partners, with regular industry steering boards and lectures to ensure content satisfies industry needs.
   Focused events (awaydays/camps) will develop generic skills (negotiation/communication/entrepreneurship), and rapid-solving of industry challenges (hackathons/living-lab). Graduating cohorts will organise and host free format workshops for other cohorts and industry partners. These generate team working and organisational abilities, training students to quickly digest and adapt knowledge to solve problems, seize opportunities and communicate to broad audiences of experts and stake holders. These skills are not normally developed through DTPs.

4. Which Schools or Faculties would be involved in either leading this CDT or supervising students in your proposed CDT?
   This CDT proposal is broad and spans a significant part of the university. We expect interaction across the faculties of Engineering and Physical Science, Environmental and Life Sciences and Social Sciences, with additional collaborations with Archaeology in the faculty of Arts and Humanities. There will be a significant spread of the senior leadership team right across the faculties to cover the core themes of the CDT and ensure that all aspects of the Geospatial Artificial Intelligence lifecycle from data acquisition to intelligent and useful decision making are represented. The breadth of engagement across the university will provide a wide pool of supervisors and prospective (funding-eligible) PhD students.

5. How does the proposed CDT align with Southampton’s strategy and the priorities of EPSRC? The CDT’s interdisciplinary nature addresses broader university strategy and builds on the idea of the triple helix of research, education and knowledge exchange around a central core of people. Nationally the proposal addresses multiple EPSRC priorities: strategic priorities of “artificial intelligence, digitisation and data” and “the physical and mathematical sciences powerhouse” are clearly at the centre. The proposal fits in the “artificial intelligence and robotics” and “digital twins” themes. More broadly, the proposal has clear application links to net zero, and security and resilience (both in terms of defence and climate change).

6. Why would a prospective PhD student want to come to this CDT? What is the unique selling point?

   • USP1 – Knowledge-broadening training:
     Recruitment from diverse ranges of backgrounds, access to tailored Masters-level modules targeting both technical aspects and the broader ethical, privacy and legal issues and professional practices surrounding geospatial data use, combined with strong links to external organisations (USP3) will make the CDT a vibrant and exciting place to study and work. The activities naturally broaden cohort-wide knowledge and build networks for future careers. We further expect CDT students will benefit from broader pre-doctoral academic activities that are being actively considered, such as the development of UG and MSc programmes that complement our unique AI and geospatial USPs and can both complement and feed into the CDT cohort.
   • USP2 – Access to the facilities and unique breadth of expertise:
     Geospatial data and analytics interfaces with all sectors, where domain specific aspects can form levers or barriers for state-of-the-art AI techniques. Industry, government, and academic groups increasingly recognising that the expertise needed to solve geospatial data challenges differ from general AI use. Although many universities are developing hubs in AI for medicine/healthcare, AI and quantum technologies, no single institution can match the critical mass UoS has in geospatial AI. UoS has strengths across all relevant disciplines, with a proven track-record in fundamental AI (ECS/Mathematics) and geospatial application domains (Engineering/Geography and Environmental Science), as evidenced by the considerable number of 4* REF2021 returns across the core Units of Assessment. These successes are enabled by facilities such as the IRIDIS 5 HPC, access to vast geospatial and temporal datasets (some unique to UoS), access to data gathering platforms, instruments and the experts who develop and deploy them.
   • USP3 – Regional enterprise networks for Geospatial AI careers and industry need:
     UoS has a strong track-record of geospatial data and analysis enterprise activities that continue to generate significant income and bring reputational benefit. Many of these have been featuring in REF2021 impact case studies (e.g. https://www.ecs.soton.ac.uk/research/novel-image-capture, https://www.southampton.ac.uk/engineering/research/impact.page, https://www.southampton.ac.uk/geography/research/impact.page), with many partner organisations based in the regional South. Regular CDT training events (awaydays/camps, hackathons, living-labs, workshops) building on these networks will facilitate deep interaction with industry partners, particularly those in our region, and present internship and career opportunities for CDT doctoral graduates.