Research
Research Theme: Energy Systems
PV Sizing and Operation
Researcher: Anaïs Berkes
Supervisor: Srinivasan Keshav
Summary:
We address the optimal sizing and operation of home energy systems integrating photovoltaic (PV) panels, electric vehicles (EVs), and home energy management systems (HEMS) in the context of increased remote working. Using our SOPEVS framework, we analyze the impact of commuting habits and HEMS preferences on system sizing. Our findings suggest that remote-working homeowners with bidirectional EVs can eliminate the need for separate home storage, reducing system costs significantly, with potential savings of up to 80% when maximizing solar energy charging.
Please click here for more details.
Research Theme: Heatwave Impact
Topic: Transmission Grid
Researcher: Enming Liang
Supervisors: Minghua Chen, Srinivasan Keshav
Summary:
The European Electricity Grid Exhibits Temperature-induced Capacity Bottlenecks During Heatwaves.
Please click here for more details.
Research Theme: Biodiversity Monitoring
Topic: LIFE/STAR
Researcher: Emilio Luz-Ricca
Supervisors: Andrew Balmford, Michael Dales, Anil Madhavapeddy, Tom Swinfield
Summary:
In response to the global biodiversity crisis and to meet demand for tools to quantify biodiversity loss or gain, members of the Cambridge Centre for Carbon Credits (4C) have developed the Land-cover change Impacts on Future Extinctions (LIFE) metric—a spatial metric that quantifies and aggregates the effect of land use changes on species extinction risk for nearly 30,000 terrestrial vertebrate species.
My work is primarily focused on improving LIFE extinction risk estimates. One aspect of this is examining the distribution and intensity of various anthropogenic threats (e.g., fragmentation, edge effects, hunting, invasive species). Another avenue for improving LIFE estimates is through the input data products; to this end, I will experiment with data-driven methods for improving species range maps with an eye towards integrating plant species into LIFE.
Please click here for more details.
Research Theme: Remote Sensing
Topic: Foundation Models for Complex Geospatial Tasks
Researcher: Onkar Gulati
Supervisors: David Coomes, Sadiq Jaffer, Anil Madhavapeddy
Summary:
Self-supervised learning (SSL) represents a shift in machine learning that enables versatile pre-trained models to leverage the complex relationships present in dense–oftentimes multispectral and multimodal–remote sensing data. This in turn can accelerate how we address sophisticated downstream geospatial tasks for which current methodologies prove insufficient, ranging from land cover classification to urban building segmentation to crop yield measurement and wildfire forecasting.
This PhD project explores the question of how current SSL methodologies may be altered to tackle remote sensing tasks, and also how to make them amenable to incremental time-series generation as new data regularly comes in from sensing instruments.
Please click here for more details.
Topic: Forest Degradation
Researcher: Amelia Holcomb
Supervisors: David Coomes, Srinivasan Keshav
Summary:
Forest disturbance, defined as a partial reduction in forest cover that does not result in conversion to non-forested land, has surpassed deforestation by area in the Brazilian Amazon. In addition to causing direct carbon emissions, disturbance also diminishes ecosystem integrity by harming forest structure, even when canopy cover remains. Recent advances using LandSat and Sentinel-1 have improved the detection of disturbances at fine spatiotemporal resolution but are so far unable to quantify the changes in forest structure and biomass associated with a detected disturbance. The Global Ecosystem Dynamics Investigation (GEDI), a novel spaceborne LiDAR system, has captured billions of 25-meter diameter footprints measuring forest height, plant area, and understory structure since it began collecting data in 2019. Though there is no guaranteed repeat cycle, GEDI often measures the same location several times; some of these coincident footprints happen to fall before and after a detected disturbance. In this work, we develop a general-purpose open-source pipeline for identifying these locations and use it to find over 7,100 coincident footprint pairs with intermediate disturbance events across the Amazon basin. We also identify a control set of over 34,000 coincident footprint pairs from disturbed areas but without intermediate disturbance events. Analysis of this continent-scale dataset demonstrates that GEDI can detect canopy and biomass losses in non-stand-replacing disturbances as small as 0.09 ha. GEDI’s unique three-dimensional view of forest structure allows us to identify varied effects of different disturbance types, including areas where the upper canopy retains its height, but the understory suffers substantial losses. Finally, we model the relationship between LandSat and Sentinel-1 disturbance detection parameters and GEDI-measured per cent biomass loss. This is the first step towards a pan-tropical, spatially explicit estimate of carbon losses and structural changes due to forest disturbance.
Please click here to view the paper.
Topic: The Role of Urban Vegetation in Human Health
Researchers: Andrea Domiter, Andrés Zúñiga-González
Supervisors: Srinivasan Keshav, Anil Madhavapeddy, Ronita Bardhan
Summary:
As cities become home to over 70% of the global population by the mid-21st century, they face unprecedented challenges in sustainability and climate regulation. This PhD project aims to model the crucial role of urban vegetation in regulating city climates and improving human health, using a data-driven approach that combines weather, remote sensing, and socio-economic data. By analyzing the implementation of green spaces and trees in urban environments, the research seeks to address the complex interplay between urban structure, environmental sustainability, and social equity in shaping the cities of the future.
Please click here for more details.
Topic: Diffusion Models for Terrestrial Predictions about Land Use Change
Supervisors: Anil Madhavapeddy, Sadiq Jaffer
Summary:
This project aims to develop and evaluate a video diffusion model trained on satellite imagery time series to predict and visualize landscape evolution, including changes in land use and cover, with potential applications in forecasting deforestation, flooding, and fire risks.
Please click here for more details.
Topic: Low-power Sensing Infrastructure for Biodiversity
Researcher: Josh Millar
Supervisors: Anil Madhavapeddy, Hamed Haddadi
Summary:
This project aims to develop a low-power, multi-sensor device using on-device reinforcement learning to optimize scheduling and cooperation between devices, enhancing operational lifetime and coverage for long-term deployment in remote environments..
Please click here for more details.
Research Theme: Forest Carbon Monitoring
Topic: Trunk Diameter
Researcher: Frank Feng
Supervisor: Srinivasan Keshav
Summary:
In forestry and ecological studies, accurately measuring tree trunk diameter is essential for monitoring forest health, estimating biomass, and conducting various environmental assessments. Traditional methods are often time-consuming and labor-intensive. While some smartphone-based tree-trunk diameter measurement techniques have been proposed in recent years, most of these solutions require high-end depth sensors.
In response to these challenges, we have developed a novel smartphone application capable of measuring tree trunk diameter without the need for specialized depth sensors, thereby expanding accessibility to both high-end and low-end smartphone users. Our app leverages advanced neural networks and image processing algorithms to accurately estimate tree trunk diameter from images taken by the optical camera.
Please click here for more details.
Topic: Impact of Fires
Researcher: Jovana Knezevic
Supervisors: David Coomes, Srinivasan Keshav, Anil Madhavapeddy
Summary:
Topic: 3D Reconstruction
Researcher: Yihang She
Supervisors: Andrew Blake, David Coomes, Srinivasan Keshav
Summary:
The research aims to develop a toolkit for creating augmented digital twins of forests to enable scalable monitoring algorithms using synthetic data, addressing the limitations of existing forest monitoring methods. The project is divided into two stages: constructing virtual forests from real forest data, and generating synthetic data from these digital twins to train 3D perception algorithms for forest property inference.
Please click here for more details.
Topic: Quantifying Carbon Sequestration in Regrowing Forests in the Brazilian Atlantic
Researcher: Felipe Begliomini
Supervisors: David Coomes, Srinivasan Keshav, Charlotte Wheeler
Summary:
The content of the research will be provided soon.
Research Theme: Computer Systems & Environment and Nature
Topic: Assessing Mangrove Literature for Conservation Evidence
Supervisors: Anil Madhavapeddy, Sadiq Jaffer and Tom Worthington
Summary:
The Conservation Evidence Copilots project aims to synthesize and evaluate diverse literature on mangrove forest conservation, integrating expert knowledge and remote sensing data to develop best practices and inform policy decisions for this threatened ecosystem.
Please click here for more details.
Topic: Spatial and Multi-modal Extraction from Conservation Literature
Supervisors: Anil Madhavapeddy, Sadiq Jaffer, Alec Christie, Bill Sutherland
Summary:
This MPhil project aims to combine published literature resources with geographic information to propose evidence-driven conservation interventions, enhancing the targeting and impact of future conservation efforts. The project involves developing a pipeline for extracting conservation-related information from literature, creating multimodal models for analyzing scientific papers, and building a predictive model to assess the potential efficacy of conservation interventions using machine learning and LLM techniques.
Please click here for more details.
Topic: Accurate Summarisation of Threats for Conservation Evidence Literature
Researchers: Kittson Hamill
Supervisors: Anil Madhavapeddy, Sadiq Jaffer
Summary:
The Conservation Evidence Copilots project aims to synthesize and evaluate diverse literature on mangrove forest conservation, integrating expert knowledge and remote sensing data to develop best practices and inform policy decisions for this threatened ecosystem.
Please click here for more details.
Topic: Generating Chunk-free Embeddings for LLMs
Researchers: Mark Jacobsen
Supervisors: Anil Madhavapeddy, Sadiq Jaffer
Summary:
The Conservation Evidence Copilots project aims to develop a reliable pipeline for extracting, synthesizing, and verifying conservation literature to create an accurate taxonomy of wildlife threats, with the goal of informing policy and practice through clear, concise, and expert-validated summaries.
Please click here for more details.
Research Theme: Planetary Computing
Topic: Composable Diffing for Heterogenous File Formats
Supervisors: Patrick Ferris, Anil Madhavapeddy
Summary:
This project aims to develop a composable diffing specification and prototype tool for comparing and merging heterogeneous geospatial data formats, using a domain-specific language to define flexible diffing rules, and evaluate its effectiveness against conventional Git-based approaches using real-world remote sensing datasets.
Please click here for more details.
Topic: Towards Reproducible URLs with Provenance
Supervisors: Patrick Ferris, Anil Madhavapeddy
Summary:
This project aims to develop a practical implementation of Vurls (versioned URIs) by integrating the concept into a popular HTTP library, creating a proxy service for resolution and version comparison, and exploring compact representations and integration with existing web infrastructure.
Please click here for more details.
Topic: Privacy-preserving Emissions Disclosure Techniques
Researcher: Jessica Man
Supervisors: Martin Kleppmann, Anil Madhavapeddy
Summary:
This PhD research investigates the development of verifiable and privacy-preserving carbon emissions reporting mechanisms across the cloud computing supply chain, aiming to enable competition based on sustainability and drive demand-side pressure for renewable energy adoption in cloud services.
Please click here for more details.
Topic: Effective Geospatial Code in OCaml 2024
Researcher: George Pool
Supervisors: Anil Madhavapeddy, Michael Dales, Patrick Ferris
Summary:
Customers of online services may want to take carbon emissions into account when deciding which service to use, but are currently hindered by a lack of reliable emissions data that is comparable across services. Calculating accurate carbon emissions across a cloud computing pipeline involves a number of stakeholders, none of whom are incentivised to accurately report their emissions for competitive reasons.
This PhD explores mechanisms to support verifiable and privacy-preserving emissions reporting across a chain of energy suppliers, cloud data centres, virtual machine hosting services providers and cloud services providers, which are ultimately passed through to APIs used by customers. We hypothesise that adding verifiable and composable emissions transparency to cloud computing architectures enables providers to compete on the basis of sustainability, resulting in demand-side pressure on cloud services to shift to renewable energy sources.
Please click here for more details.
Topic: Gradually Debugging Type Errors 2024
Researcher: Max Carrol
Supervisors: Patrick Ferris, Anil Madhavapeddy
Summary:
This project aims to enhance OCaml's type error reporting by incorporating gradual typing, allowing for a more intuitive and exploratory approach to understanding and resolving type errors, especially in larger codebases.
Please click here for more details.
Topic: An Imperative, Pure and Effective Specification Language 2024
Researcher: Max Smith
Supervisors: Patrick Ferris, Anil Madhavapeddy
Summary:
This project aims to develop a Python-like specification language with typed holes and functional underpinnings, translating to Hazel or OCaml 5 for robust semantics and multiple execution modes, including interactive tracing and high-performance parallel processing.
Please click here for more details.
Topic: Computational Models for Scientific Exploration
Researcher: Patrick Ferris
Supervisor: Anil Madhavapeddy, Srinivasan Keshav
Summary:
This research project addresses the computational challenges in climate science and ecology by conducting a systematic study of uncertainty sources in these fields. The goal is to develop a specification language and hermetic computation environment that enables scientists to create more precise, testable methodologies while maintaining the flexibility to explore intermediate results, ultimately advancing the scientific method in computational research.
Please click here for more details.
Topic: Compressive Streaming for Geospatial Pipelines
Researcher: Omar Tanner
Supervisor: Anil Madhavapeddy, Sadiq Jaffer
Summary:
This project enhances geospatial data processing by using modern lightweight compression techniques to improve CPU and RAM bandwidth efficiency, leveraging CPUs' superscalar capabilities and SIMD instructions for faster data access. By employing multi-objective Pareto optimization, it achieves significant speedups in data management and analysis, crucial for environmental challenges, and provides a benchmark for optimizing geospatial pipelines.
Please click here for more details.
Research Theme: Mapping LIFE on Earth
Topic: Mapping Hunting Risks for Wild Meat in Protected Areas 2024
Researcher: Charles Emogor
Supervisor: Anil Madhavapeddy, Milind Tambe
Summary:
This project aims to optimize anti-poaching ranger patrols and address knowledge gaps in wild meat snaring using machine learning techniques and novel base maps, balancing biodiversity protection with local community well-being in protected areas.
Please click here for more details.
Topic: Using wasm to Locally Explore Geospatial Layers 2024
Researcher: Sam Forbes
Supervisor: Michael Dales, Anil Madhavapeddy
Summary:
This project aims to develop a WebAssembly-based visualization tool for geospatial ecology data, enabling efficient, in-browser processing and display of species distribution maps from GeoTIFF files, with a focus on user-specified subsets and concurrent processing for real-time performance.
Please click here for more details.
Topic: Real-time Mapping of Changes in Species Extinction Risks 2024
Researcher: Emilio Luz-Ric
Supervisor: Andrew Balmford, Anil Madhavapeddy
Summary:
This PhD project aims to develop advanced modeling and machine learning techniques that integrate remote sensing data with other information sources to assess and quantify the impacts of complex anthropogenic threats, beyond simple habitat loss, on species-specific habitat quality and occupancy.
Please click here for more details.
Interspatial OS
Topic: Deep Learning for Decomposing Sound into Vector Audio 2024
Researcher: Trevor Agus
Supervisor: Andrew Balmford, Anil Madhavapeddy
Summary:
This project aims to use deep learning techniques to decompose complex recorded sounds into perceptually relevant components based on cochlear activity patterns, potentially advancing hearing research and applications in sound processing.
Please click here for more details.
Topic: Low-latency Wayland Compositor in OCaml 2024
Researcher: Tom Thorogood
Supervisor: Ryan Gibb, Anil Madhavapeddy
Summary:
This project aims to develop a Wayland compositor in OCaml, leveraging the language's safety and composability features to create customizable window management logic for next-generation display servers, potentially revolutionizing the approach to situated displays and hybrid streaming systems.
Please click here for more details.
Topic: Building Bigraphs of the Real World 2024
Researcher: Roy Ang
Supervisor: Ryan Gibb, Anil Madhavapeddy
Summary:
This project aims to create a comprehensive bigraph model of the physical world using OpenStreetMap data, integrating hierarchical place structures and street connectivity, to enhance location-aware applications and demonstrate its utility in modeling ubiquitous systems like Bluetooth connectivity.
Please click here for more details.
Topic: Interspatial Networking with DNS
Researcher: Ryan Gibbs
Supervisor: Anil Madhavapeddy, Jon Crowcroft
Summary:
This PhD project explores the Spatial Name System (SNS), a novel approach to addressing the limitations of the current Internet architecture in naming and resolving location-based services. By extending the DNS to incorporate hierarchical location-based names, SNS aims to enable both global and local resolution schemes, integrating spatial names into existing applications and opening new possibilities for sensor networks and augmented reality.
Please click here for more details.
EEG Group Research Archives
