A Robust Software Stack for Omnidirectional Mobile Robots: From Simulation to Deployment in Real-World Environments

This thesis suggests the development of a complete software stack for an autonomous robot, from simulation to deployment into the world. The project, conducted with the Hilti Research and Development Robotics department, had as its goal giving the robot the capability to autonomously travel from designated waypoints, not bumping into anything while maintaining high accuracy. The general goal was the provision of a functioning demonstration proving the capability of the system to stakeholders on behalf of the company. The project was structured into several significant phases. The robot was initially modelled in URDF and simulated in Gazebo to create a correct digital twin for testing. ROS2 Control was employed to develop a solid motion control system with the capability to control velocity smoothly and robustly. The odometry pipeline was additionally tuned to enhance the accuracy of localization. Two LiDAR sensors were placed strategically, and their point clouds were merged and filtered for increased perception at the expense of lower computational load. Additionally, Zenoh middleware was introduced for more fault-tolerant communication infrastructure. One of the most critical aspects of this endeavour was to fine-tune the SLAM and navigation stack in simulation as well as real-world environments so that it would provide safe, collision-free navigation through dynamic spaces. The autonomous navigation system was planned to navigate narrow corridors while providing accuracy, and safety protocols were put in place to avoid bumping into moving objects or humans. The final demonstration successfully showcased the robot navigating complex environments with high precision, efficiently reaching target positions while avoiding obstacles. The developed software architecture proved to be reliable and effective, generating strong interest among Hilti’s stakeholders. Although the project was completed as planned, a potential future improvement could involve integrating Hilti’s total station to further enhance positioning accuracy using an onboard prism. This thesis formulates the issues and solutions of creating a scalable and modular software architecture for autonomous mobile robots. The insights from modelling, control, sensor fusion, and navigation in the ROS2 environment are also applicable to the broader field of robotic autonomy and industrial automation.