Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots

TL;DR

A fully integrated system comprising adaptive locomotion control, mobility-aware local navigation planning, and large-scale path planning within the city is introduced, supporting the feasibility of wheeled-legged robots and hierarchical RL for autonomous navigation, with implications for last-mile delivery and beyond.

Abstract

Autonomous wheeled-legged robots have the potential to transform logistics systems, improving operational efficiency and adaptability in urban environments. Navigating urban environments, however, poses unique challenges for robots, necessitating innovative solutions for locomotion and navigation. These challenges include the need for adaptive locomotion across varied terrains and the ability to navigate efficiently around complex dynamic obstacles. This work introduces a fully integrated system comprising adaptive locomotion control, mobility-aware local navigation planning, and large-scale path planning within the city. Using model-free reinforcement learning (RL) techniques and privileged learning, we develop a versatile locomotion controller. This controller achieves efficient and robust locomotion over various rough terrains, facilitated by smooth transitions between walking and driving modes. It is tightly integrated with a learned navigation controller through a hierarchical RL framework, enabling effective navigation through challenging terrain and various obstacles at high speed. Our controllers are integrated into a large-scale urban navigation system and validated by autonomous, kilometer-scale navigation missions conducted in Zurich, Switzerland, and Seville, Spain. These missions demonstrate the system's robustness and adaptability, underscoring the importance of integrated control systems in achieving seamless navigation in complex environments. Our findings support the feasibility of wheeled-legged robots and hierarchical RL for autonomous navigation, with implications for last-mile delivery and beyond.

Figures (9)

• Figure 1: Deployments in urban environments. Our control system for the wheeled-legged robot has undergone extensive validation in various indoor and outdoor locations. The experiments took place in Zurich, Switzerland and in Seville, Spain. (A) Locomotion challenges. (B) Navigation challenges; dynamic and static obstacles, complex terrains, and narrow space. (C) Locations in Zurich. (D) Locations in Seville.
• Figure 2: System overview.(A) Our wheeled-legged quadrupedal robot is equipped with various payloads for onboard terrain mapping, obstacle detection, and localization. (B) Overview of the navigation system. The system is driven by two neural network policies operating at different levels. The high-level navigation policy observes two waypoints (WP 1 and WP 2) and generates target velocity commands for the locomotion policy. The low-level locomotion policy then controls joint actuators and follows the velocity commands. (C) Our training environment is designed to dynamically generate new navigation paths for each episode, optimizing the learning process. By leveraging pre-generated obstacle-free paths, we enhance the navigation capabilities of our system.
• Figure 3: Large scale autonomous navigation experiment at Glattpark, Zurich.(A) Our city navigation workflow begins with offline preparation, involving scanning the test area using a handheld laser scanner and constructing a navigation graph. (B) The robot autonomously navigated the urban environment to reach 13 predetermined goal points, selected in an arbitrary order. (B-i, ii) Path planning within the city was facilitated by the pre-generated navigation graph. (B-iii) Moving speed and mechanical cost of transport compared to a normal legged robot (ANYmal-C).
• Figure 4: Challenges in the populated urban environment.(A) The urban environment presents various obstacles. Some have to be avoided, such as pedestrians or poles, and others can be traversed, such as stairs or steps. (B) We had to intervene and stop the mission in these three cases.
• Figure 5: Obstacle negotiation. (A) Our robot navigates around blocked routes by actively exploring the area and finding alternative paths. (B) Safe traversal of a narrow space. (C) Our robot exhibits two different ways to traverse the complex obstacle. (C, D) Our robot shows an asymmetric understanding of traversability, being able to traverse higher steps when going down. (D) We ensure safety around humans by incorporating additional human detection and overriding height scan values.