Details
This research project focuses on the development of methods for intelligent wildfire detection and localisation using swarms of unmanned aerial vehicles (UAVs). Current state of the art methods, which usually rely on optical or infrared (IR) cameras, are seldom resilient and often trigger false positive detections. To fill in this gap, in collaboration with industrial partners, the research will develop novel Machine Learning and Computer Vision methods for detecting and localising. These will be used to develop automatic and/ or human in the loop guided decisions for the SWARM of unmanned aircraft, enabling greater speed of response for firefighting teams. The project aims to develop real-time solutions for a small, medium and large UAVs operated independently or as a swarm.
This research focuses on the development of methods for intelligent sensing with swarms of unmanned aerial vehicles (UAVs) for autonomous wildfire detection and localisation. Currently there are no well-working solutions, since most developed methods with optical cameras face a significant number of false positive detections. To fill in this gap, in collaboration with industrial partners, the research will develop novel Machine Learning and Computer Vision methods for wildfire detection, localisation and decision making to support firefighting search and rescue teams. The project scope is on real-time solutions for a small, medium and large number of entities with a swarm of UAVs.
Computer Vision and Machine Learning approaches will be developed and combined with regression methods and provide robust and trustworthy solutions to lighting conditions, occlusions and to other environmental changes.
The robustness of the developed approaches will be studied both theoretically and experimentally, in collaboration with industrial partners. Hence, the development of approaches that provide an increased level of autonomy on centralised and distributed tasks such as inspection of a big geographic area for fires is an important objective of this project.
The project aims are to:
1) provide methodological contributions towards intelligent sensing with UAVs for tackling wildfires, including modular, scalable methods with respect to the volume of collected data, to the number of UAVs, data sampling rates, number of sensors, centralised, decentralised methods and other factors. Image detection and segmentation, dynamic colour models for fire appearance in image and fusion of image features are in the project scope. Different types of images such as optical, IR and hyperspectral will be considered.
2) quantify the impact of uncertainties with respect to image noises, environmental changes, test, validate, evaluate the developed approaches both in simulated environments and with real data on real UAVs. Defining and calculating measures for levels of trust in the developed algorithms is essential. These uncertainty-aware algorithms can self-assess performance and continually learn from the available datasets while operating in real-time on a limited energy budget.
3) develop methods for UAV autonomous landing, without any pilot from the ground, in different weather conditions.
4) develop methods for sensor management and data fusion linked with inference and decision making, jointly applied to several wildfire detection tasks.
5) embed the developed approaches within a generic Digital Twin framework provided by the industrial partners and validate them on real UAVs in combination with the UAV navigation system and disseminate the results via publications.
There are opportunities for industrial placements during the project.
The PhD student will be based at the School of Electrical and Electronic Engineering at the University of Sheffield.
The School of Electrical and Electronic Engineering is world-leading and with the quality of our research environment rated as top nationally. The University of Sheffield is a vibrant, innovative and supportive place to undertake research.
For additional information, please feel free to contact Prof. Lyudmila Mihaylova( l.s.mihaylova@sheffield.ac.uk ) with your CV and relevant qualifications /transcripts.
Interested candidates are strongly encouraged to contact the project supervisors to discuss your interest in and suitability for the project prior to submitting your application.
We require applicants to have either an undergraduate honours degree (1st/2:1) or MSc (Merit or Distinction) in a relevant science or engineering subject from a reputable institution.