This paper presents a comprehensive system for autonomous navigation and locomotion in wheeled-legged robots, designed to enhance logistics systems and improve operational efficiency in urban environments. The system integrates adaptive locomotion control, mobility-aware local navigation planning, and large-scale path planning within cities. Using model-free reinforcement learning (RL) and privileged learning, the authors develop a versatile locomotion controller that can efficiently navigate various terrains and obstacles. The controllers are integrated into a hierarchical RL framework, enabling effective navigation through challenging terrain and obstacles at high speeds. The system was validated through kilometer-scale autonomous navigation missions in Zurich, Switzerland, and Seville, Spain, demonstrating robustness and adaptability. The research highlights the potential of wheeled-legged robots and hierarchical RL for autonomous navigation, with implications for last-mile delivery and beyond. Key contributions include the development of a robust locomotion controller, a mobility-aware navigation controller, and the integration of these components into a seamless autonomous navigation system. The system's performance is compared with conventional navigation approaches, showing superior success rates and computational efficiency. Future work will focus on enhancing perception capabilities and reducing human labor in map creation.This paper presents a comprehensive system for autonomous navigation and locomotion in wheeled-legged robots, designed to enhance logistics systems and improve operational efficiency in urban environments. The system integrates adaptive locomotion control, mobility-aware local navigation planning, and large-scale path planning within cities. Using model-free reinforcement learning (RL) and privileged learning, the authors develop a versatile locomotion controller that can efficiently navigate various terrains and obstacles. The controllers are integrated into a hierarchical RL framework, enabling effective navigation through challenging terrain and obstacles at high speeds. The system was validated through kilometer-scale autonomous navigation missions in Zurich, Switzerland, and Seville, Spain, demonstrating robustness and adaptability. The research highlights the potential of wheeled-legged robots and hierarchical RL for autonomous navigation, with implications for last-mile delivery and beyond. Key contributions include the development of a robust locomotion controller, a mobility-aware navigation controller, and the integration of these components into a seamless autonomous navigation system. The system's performance is compared with conventional navigation approaches, showing superior success rates and computational efficiency. Future work will focus on enhancing perception capabilities and reducing human labor in map creation.