Vision-based Navigation for Soft Endoluminal Robots

Project ID: 2228cd1291 (You will need this ID for your application)

Research Theme: Artificial Intelligence and Robotics

Lead Supervisor: Sophia Bano

Project Summary:

Significance of the Research: Endoluminal interventions (e.g. endoscopy, otoscopy) are minimally invasive procedures for the disease diagnosis and treatment. Limited field-of-view, constrained manoeuvrability of the endoscope, scene texture paucity and poor video feedback during such interventions hinder the complete inspection of the endoluminal environment, leading to missed diagnosis and treatment. To improve the dexterity and precision during these interventions, soft robots are being developed which are yet to be equipped with artificial intelligence capacities to allow real-time accurate navigation. Unlike real-world mobile robot navigation, where multitude of sensors (multiple cameras, LiDAR, GPS, IMU, etc) are available, a monocular miniature camera is the only sensing source making the endoluminal robot navigation problem non-trivial. The problem will focus on developing vision-based robot localisation and mapping techniques, enabling complete inspection of the endoluminal environment.

Collaborative Environment: This project will take place at UCL WEISS’ Surgical Robot Vision group, experts in Surgical AI and Robotics. We collaborate closely with surgeons to develop advanced AI-driven solutions, holding weekly one-on-one meetings for personalized support and regular team meetings to track progress and ensure technical advancement.

Research Objectives: The primary objectives of this project include:

Designing and developing deep learning techniques for vision-based robot pose and camera depth estimation.
Creating a robust localization and 3D reconstruction system for endoluminal environments using only data from the onboard monocular camera.
Integrating the developed navigation system into real-time robot control systems through close collaboration with the robotics team.
Exploring additional research avenues, such as endoluminal scene understanding, mapping, and reconstruction.

Desired Qualifications: We are seeking highly motivated candidates with the following qualifications:

Undergraduate degree in computer science, biomedical engineering, mechatronics, or related fields.
Strong knowledge of machine learning, computer vision, and/or artificial intelligence.
A passion for advancing medical technology through interdisciplinary research.