Relative Localization, and Simultaneous Localization and Mapping

Abstract

Relative localization is essential for a robot to perceive what lies within its immediate environment, and also to help it navigate. In the past, robot systems have employed many technologies to perform relative localization, including various motion detection systems as well as Simultaneous Localization and Mapping techniques. With the reduced cost of cameras with good resolution and increased computational capabilities of robotic on-board computers, computer vision techniques are now being investigated as additive sensors to inertial measurement unit systems. Most computer vision techniques employ markers, called fiducial markers, that have associated algorithms that perform image correction and pose estimation. My research involved using visual servoing with the Open Manipulator robot and the oCAM to perform relative bearing and distance measurements of fiducial markers. The results were validated using a Vicon system with inverse transformations. My results indicated that computer vision techniques were very effective when ambient lighting conditions were controlled.

My Work

• Detected and identified fiducial markers with oCAM cameras and performed oCAM calibration with Charuco.
• Performed image operations (edge and corner detection) using OpenCV libraries.
• Performed relative bearing and distance measurements of fiducial markers using visual servoing with the Open Manipulator robot and the oCAM.
• Conducted experiments to validate the results from the Computer Vision system using a Vicon system.