Skill-Nav: Enhanced Navigation with Versatile Quadrupedal Locomotion via Waypoint Interface

TL;DR

Skill-Nav introduces a hierarchical navigation framework for quadrupedal robots that integrates reinforcement learning-based locomotion skills with waypoint-guided control, enabling effective traversal of complex terrains.

Contribution

The paper presents a novel waypoint interface for quadrupedal locomotion, combining RL-trained policies with high-level planning tools for improved navigation.

Findings

Effective terrain traversal demonstrated in simulation and real-world

Waypoint interface simplifies high-level planning and control

Enhanced navigation capabilities over complex terrains

Abstract

Quadrupedal robots have demonstrated exceptional locomotion capabilities through Reinforcement Learning (RL), including extreme parkour maneuvers. However, integrating locomotion skills with navigation in quadrupedal robots has not been fully investigated, which holds promise for enhancing long-distance movement capabilities. In this paper, we propose Skill-Nav, a method that incorporates quadrupedal locomotion skills into a hierarchical navigation framework using waypoints as an interface. Specifically, we train a waypoint-guided locomotion policy using deep RL, enabling the robot to autonomously adjust its locomotion skills to reach targeted positions while avoiding obstacles. Compared with direct velocity commands, waypoints offer a simpler yet more flexible interface for high-level planning and low-level control. Utilizing waypoints as the interface allows for the application of various general planning tools, such as large language models (LLMs) and path planning algorithms, to guide our locomotion policy in traversing terrains with diverse obstacles. Extensive experiments conducted in both simulated and real-world scenarios demonstrate that Skill-Nav can effectively traverse complex terrains and complete challenging navigation tasks.