Dynamic Urban

Dynamic urban modeling focuses on accurately representing and predicting the complex, ever-changing nature of urban environments, primarily for applications in autonomous navigation and scene understanding. Current research emphasizes developing efficient and robust methods for reconstructing 3D urban scenes, including dynamic objects, often leveraging neural radiance fields (NeRFs) and Gaussian splatting techniques, along with ensemble learning and spatio-temporal prediction networks. These advancements are crucial for improving the safety and efficiency of autonomous vehicles and other robotic systems operating in urban areas, as well as for creating more realistic and detailed virtual urban environments for simulation and training. The development of large, diverse datasets is also a key focus to enable the training and evaluation of these advanced models.