Research Themes

The SustAI CDT focuses around 5 research themes where AI can improve environmental sustainability:

AI for Transportation and Logistics

Theme Leads: Professor Selin Ahipasaoglu and Professor Sebastian Stein
This theme focuses on using AI to optimise transportation and logistics operations for sustainability, including reducing energy and resource use, improving smart mobility services, and developing more sustainable transportation systems. 

Example applications include intelligent intersection management for connected and autonomous vehicles; ride sharing and dynamic bus routing; micro-tolling to reduce congestion; personalised electric vehicle charging and routing; optimising shipping routes and fleet management using digital twins.

AI for Sustainable Energy and Buildings

Theme Leads: Dr Stephanie Gauthier and Professor Ajit Nayak
This theme focuses on using AI to optimise and improve the management of sustainable energy technologies such as solar, wind, and hydro, including forecasting and predicting energy demand, improving efficiency and reducing costs, smart grid and energy markets, and managing energy use in smart buildings. 

Examples include optimising energy use and storage based on occupancy prediction and price forecasts; creating novel community energy markets to efficiently use local production; understanding and managing the trade-offs between cost and comfort in homes and offices. Human-in the loop learning approaches which minimise human participation and are explainable.

AI for Sustainable Operations and Circular Economy

Theme Leads: Dr Vahid Yazdanpanah and Dr Hector Calvo-Pardo
This focuses on using AI to optimise manufacturing processes, operations, and supply chains for sustainability, including reducing waste and energy use, improving resource efficiency, waste management and recycling, and developing more sustainable products, packaging, and service delivery models within a circular economy framework. A key aspect is using AI to develop and evaluate appropriate policies and incentives, including green finance and carbon emission trading schemes.

Example uses of AI include anomaly detection for improved maintenance; prediction of demand and supply enabling more efficient supply chains and reduced waste; tracking of CO2 throughout the supply chain to incentivise behavioural change; predicting and mapping particle emissions in urban areas; study of incentives and policies using game theory and agent-based modelling.

AI for the Natural Environment

Theme Lead: Professor Lindsay-Marie Armstrong
This theme focuses on using AI to protect and conserve biodiversity, including monitoring and predicting changes in ecosystems, identifying and mitigating threats to endangered species, and developing sustainable agricultural, fisheries and land management practices. Other aspects include air pollution monitoring and developing interventions.

For example, AI can process geodata and optimise resource use through data-driven modelling of the complex interactions within ecosystems; it can help conservation through identification and counting of species from audio and video streams; specialised ground , air or underwater robotics can assist monitoring and management of land and seas. Typical AI techniques include machine vision as well as robotics, UAVs, UUVs and swarm robotics. 

Sustainable AI

Theme Lead: Dr Dimitra Georgiadou
Whilst the other 4 themes focus on applications of AI, this theme focuses on reducing the power consumption that is associated with the use of AI itself. It investigates both creating more efficient machine learning algorithms to be used, e.g., using edge computing approaches, as well as introducing a fundamentally different approach at the hardware level, such as new neuromorphic architectures that emulate the synaptic plasticity of the brain.

Example projects include developing nano-electronic technologies using the University of Southampton’s world-leading clean room facilities; employing hardware acceleration for real-time AI; as well as developing novel algorithmic techniques, such as deep learning, Bayesian inference and optimisation for low-power devices.