Simulation and implementation of SLAM based on ROS 2 mobile robot

• Main Author: Loh, Xin Zhi
• Other Authors: Chen I-Ming
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Engineering; Robotics; SLAM; Simulation
• Online Access: https://hdl.handle.net/10356/177217
• Institution: Nanyang Technological University

Interest in Simultaneous Localisation and Mapping (SLAM) technology is gaining traction due to the advent of autonomous mobile systems like driverless vehicles and humanoid robots. Visual SLAM (VSLAM) using cameras is increasing, becoming the prefer approach for SLAM over 2D or 3D LiDAR SLAM, despite the latter being more mature. This can be attributed to the versatility of using vision sensors to navigate through complex environment and perform highly intelligent tasks. For developers to develop SLAM / VSLAM applications, robotic simulators can be leveraged to test programs virtually without incurring cost from procuring expensive robots or components. However, simulation results may not always be reliable even with highly accurate physics engine. This study explores the implementation of SLAM / VSLAM in NVIDIA Isaac Sim simulations, with Robot Operating System (ROS) 2 as the robotic framework and RTAB-Map chosen as the SLAM framework. Firstly, 2D LiDAR SLAM was performed on the Turtlebot3 in simulation and real environment, where mapping results were compared to highlight any gaps between them. Next, 2D LiDAR SLAM and RGBD VSLAM was conducted on the Nova Carter mobile robot in simulation to highlight the benefits of 3D VSLAM in perceiving more environmental details. Lastly, to test the capabilities of VSLAM navigation in complex environments and scenarios, a mobile manipulator navigation task was designed and the algorithm to generate the optimal waypoints to navigate its base was developed. Findings in this study will showcase the capabilities and limitations of modern robotic simulators for SLAM and VSLAM, and the versatility in designing the simulation environment for testing.