Weighted linear point-cloud registration for scan-matching in GPS-denied environments

The ability for an autonomous robot to create its own map based on onboard sensors and simultaneously, localize itself within this map is know as Simultaneous Localization and Mapping (SLAM). Although the theory behind SLAM has been well developed much work still needs to be done in realizing SLAM solutions that meet situation-specific real-world requirements. This is because sensors and actuators onboard a robot are always corrupted by noise. In particular, Unmanned Aerial Vehicles (UAVs), which travel in 3D, face additional difficulties due to the nonlinearities associated with rotation. Furthermore, indoor autonomous navigation is particularly challenging without a reliable inertial measurement for pose correction. The focus of this research project is not only to develop a robust and accurate 3D SLAM method to be used online by UAVs but to also achieve this indoors. This project will not only advance the field of autonomous indoor navigation but will also help ARA Robotique to be competitive in the UAV market.

Intern: 
Duowen Qian
Faculty Supervisor: 
James Forbes
Province: 
Quebec
University: 
Partner: 
Partner University: 
Program: