MuJoCo uses MJCF (MuJoCo XML Format) as its scene description language, which includes elements such as worldbody, actuator, sensor, and tendon; compared to URDF, it is more expressive and supports soft bodies and musculoskeletal models.