Development And Testing Of A Mobile Manipulation Robot

Abstract

In this project, we present the development and testing of Libot, a mobile manipulation robot, designed specifically for automating tasks in library environments. Libot combines the functionalities of the UR5 robotic arm and MiR250 mobile platform, enhanced with a Realsense 435i depth camera, to autonomously navigate and interact within a library setting. The core of Libot’s operation is a sophisticated book detection system using the YOLO v8 deep learning model, which recognizes book spine labels from RGB images to facilitate effective book cataloging. Additionally, the robot employs a Grasp Pose Detection (GPD) algorithm integrated with the MoveIt platform to optimize book handling through precise pick-and-place operations. Enhanced with Simultaneous Localization and Mapping (SLAM) techniques, Libot proficiently adapts to dynamic library layouts, improving its navigation and operational efficiency. The system architecture, built around the Robot Operating System (ROS), ensures robust performance in handling routine library tasks and emergent scenarios without disruptions. We tested various functions in the Gazebo environment, demonstrating the system’s high 92.3% object detection accuracy, better grasping success rate and navigation capability. This study not only demonstrates the feasibility of deploying mobile manipulation robots in public service environments but also provides insights into the potential enhancements for interactive robotic systems in complex indoor spaces.