Keywords

SLAM, autonomous navigation, quadruped robot, 3D reconstruction, Unitree Go2

Abstract

This thesis develops and evaluates a ROS2-based application that enables the quadruped robot Unitree Go2 to simultaneously perform autonomous navigation, real-time mapping, and 3D reconstruction in indoor environments using Simultaneous Localization and Mapping (SLAM). The robot system, equipped with onboard sensors, aims to generate high-resolution point clouds that provide detailed 3D scans. The assessment focuses on the platform’s practical feasibility, operational reliability, and accuracy. The hardware setup combines the Unitree Go2 robot with its built-in sensors, a high-resolution 3D LiDAR, and an RGB-D camera. Software components include ROS2, the Unitree SDK, a navigation framework, and SLAM approaches. The application was developed using ROS-compatible open-source repositories for the sensors and SLAM algorithms, as well as Unitree’s components and my customizations. Tests were conducted primarily in a university hallway and evaluated the quality of 3D reconstruction, computational load, wireless control capabilities, and navigation performance. The results show that the system generates dense, accurate 3D point clouds in real time, although mapping and reconstruction inaccuracies increase with longer or complex scans. Autonomous navigation based on 2D SLAM maps proved effective in simple environments but was sensitive to sensor quality. Running 2D and 3D SLAM simultaneously with live visualization requires high computational effort, risking degraded output without powerful external processing. Wireless operation is possible over short-range, high-bandwidth networks, but is limited under unstable conditions. The use of open source frameworks such as ROS2 and established SLAM algorithms facilitates application development but requires precise adaptation to the robot and the sensors used. The developed system demonstrates practical potential for real-world tasks such as indoor exploration and reconnaissance. Future work should focus on implementing multi-sensor fusion SLAM, enhancing wireless communication, and integrating advanced navigation strategies. Combining multiple mobile robots could further enhance comprehensive mapping across diverse environments.

Completion Date

2025

Semester

Summer

Committee Chair

Reiners, Dirk

Degree

Master of Science (M.S.)

College

College of Engineering and Computer Science

Department

School of Modeling, Simulation and Training

Format

PDF

Identifier

DP0029627

Language

English

Document Type

Thesis

Campus Location

Orlando (Main) Campus

Download

Share

COinS