Marine robotics: Unmanned systems for marine litter detection and collection

Unmanned Systems research on long term autonomy is essential to develop scientific and automatic technological approaches that can lead to the detection and capture of marine litter in coastal and shallow water areas.  The PhD candidate will investigate artificial intelligent methods e.g. machine learning, deep learning using multi-modal sensors i.e visible cameras and sonar that can provide long term autonomy capabilities such as but not limited to: visual scene understanding, navigation, localization, semantic mapping to unmanned systems that are being utilized in the detection and capture of marine litter in oceans and coastal areas.


Candidates
AIR Centre PhD grants are aimed at applicants enrolled or that comply with the requirements to enrol for PhD related studies and who wish to carry out research towards this degree. Candidates with MSc. degree in Electronic and Computer Engineering (or equivalent/similar area) may apply. Scientific Publications in the topic are a plus.

Summary of work plan
The main goal is to develop novel deep learning (DL) methods for robotic perception and navigation pipelines. Develop methods able to leverage multi-sensor data for achieving a persistent and robust trajectory estimation even in the face of challenging underwater scenarios.
To achieve this goal, the research plan is divided into the following objectives:
Study, research and develop a deep learning approach to integrate multibeam acoustics/side-scanning sonar data within a DL framework.
Conduct a thorough analysis of acoustic imaging data processing pipeline and identify how to better use this data for egomotion estimation.
Conceptualize and develop a novel deep learning visual-based egomotion algorithm, integrating DL techniques for faster and more accurate model convergence.
Develop a sensor fusion DL architecture accounting for temporal dynamics. The target goal is to output an egomotion estimate affected by both visual, and acoustic imaging information.
Evaluate and benchmark the developed methods, comparing with classical state-of-the-art machine learning approaches, in real application scenarios (i.e. underwater unmanned systems use for detection and classification of marine litter).

Name(s) of supervisor(s)
Supervisor: Hugo Silva, INESC TEC
Co-Supervisor: Sen Wang, Heriot Watt University

Name(s) of hosting institution(s)
The research will be conducted at INESC TEC, in cooperation with Heriot Watt University and UFBA – Federal University of Bahia

Identification of PhD program
PhD Programme in Electrical and Computer Engineering (PDEEC), offered by the University of Porto

Please check programme’s website for PhD application deadlines here.

Jury Composition
Eduardo Silva (Chairman), INESCTEC, Portugal
Pere Ridao, Univesrsitat di Girona, Italy
Matteo Matteucci, Politecnico di Milano, Italy

FCT application form
Download and fill in the FCT Form, print it, sign it and then upload it to the Application Form where requested.
You can download the FCT form here.

GDPR Policy Terms
Before proceeding, read the declaration below with our GDPR Policy Terms. You will be asked to give your consent before submitting your application.
Read English version here.
Read Portuguese version here..

Please login or signup to continue!