Husky – Vision-based motion estimation using a mobile robot

Client: The University of Edinburgh

Date: 27th August 2016

Categories: Robotics / Software Engineering / Vision

In this thesis we compare multiple approaches for producing an accurate motion estimate. We provide comprehensive analyses of the performance of an adapted LIDAR Odometry algorithm in comparison with a Visual Odometry system. Previous work fails to perform in-depth analyses between the two techniques. Moreover, we modified an algorithm proposed by Zhang and Singh for LIDAR odometry and compared it with a Visual Odometry framework. Our experiments were conducted by teleoperating a differential drive robot to traverse a trajectory indoors and outdoors. Our research provides basis for assembling a semi-autonomous Simultaneous Localisation and Mapping (SLAM) system.

Full thesis: here.

Video: here.

Leave a Comment