This talk introduces the latest development trends in spatial AI-based autonomous navigation technologies for autonomous vehicles, such as mobile robots, legged robots, and drones. We will present real-world examples developed by the KAIST Urban Robotics Lab and introduce various spatial AI frameworks that fuse diverse sensors, including LiDAR, camera, and IMU, specifically designed for dealing with challenging environments. Additionally, the deep reinforcement learning-based blind locomotion technology called DreamWaQ and DreamWaQ++ will also be introduced.