Development of a Virtual Environment using 3D Gaussian Splatting for Learning Robot Perception at Higher Education

Educational Robotics (ER) focuses on developing tools to enhance learning experiences. Simulators are recurrent in ER due to their convenience as substitutes for physical setups, providing assistance when teaching real robotics applications. However, one limitation is often the achievable level of realism. At Tecnologico de Monterrey, students majoring in robotics learn about autonomous robots through integration projects during their junior and senior years. To expand the availability of the required test courses, we propose a scene generation pipeline to create affordable simulated environments. We developed two versions of a simulator featuring a photorealistic 3DGS environment, LiDAR signals, and a ROS2 interface. We validated our proposal through trial runs with both real and simulated robots, as well as an assessment of image quality.