Implementation of Semantic SLAM on a Mobile Manipulator System

Abstract

In the field of robotics, the development of household robots capable of performing everyday tasks continues to be a major area of research and practical interest. Many domestic chores—such as picking up and moving objects from one location to another—have been successfully performed by stationary robotic manipulators paired with visual perception systems. However, accomplishing more complex, varied, and spatially distributed tasks in real-world home environments requires a mobile platform with a more human-like form factor. These tasks demand greater flexibility, spatial awareness, and interaction capabilities than fixed systems can typically provide. This work focuses on the RBY1 robot from Rainbow Robotics, a humanoid platform designed to support advanced manipulation and mobility. A range of tools and modules were developed to enhance its functionality, including software for semantic perception, task execution, and environment interaction. This thesis provides a technical overview of these tools, highlighting their roles in collecting new datasets that can be used for semantic SLAM research. In the future, these tools can enable the robot to operate more effectively in domestic settings, towards the ultimate goal of enabling more capable home-assistive robots.