Robust monocular visual-inertial localization and mapping
Project Description

Monocular visual-inertial systems, which includes only a camera and a low cost inertial measurement unit, is one of the most compact and cost-effective sensing options for robobics and augmented reality applications. This sensor setup is widely available in almost all modern smart phones, and is being deployed for autonomous drones and self-driving cars. In this project, we aim to push the frontier of monocular visual-inertial simultaneous localization and mapping (SLAM), by extending and improving our popular open source VINS-Mono system:

The goal is to develop VINS-Mono into the most robust and versatile SLAM system for a large range of applications running on different computing platforms.

SHEN Shaojie
Course type
Applicant's Roles

* Understand methods and implementations of our existing monocular visual-inertial SLAM system.
* Contribute to the continuing development of the system, including robust large loop closure, rolling shutter compensation, online sensor calibration, etc.
* Porting of our system to Android and iOS mobile devices
* Work together with PG students to conduct challenging experiments.

Applicant's Learning Objectives

* Learn state-of-the-art methodologies on vision processing pipeline, inertial navigation, sensor fusion, and simultaneous localization and mapping
* Learn the methods and implementations of our state-of-the-art VINS-Mono system.
* Learn and practice the methodologies for large-scale software development.

Complexity of the project