Classical Autonomous Navigation for a Biped Robot Using ROS2

Abstract

This paper presents the implementation and experimental validation of a classical autonomous navigation framework adapted to the Unitree G1 biped robot. The proposed system integrates LiDAR-based perception, occupancy grid maps, probabilistic localiza tion, global path planning, and velocity-based trajectory following developed within the ROS2 framework. A static map of the environment is used as the main spatial representation, while Adaptive Monte Carlo Localization estimates the robot pose with respect to the map. A global path planner based on the A* algorithm computes collision-free routes between the initial robot position and a user-defined goal. The resulting path is followed by generating velocity commands compatible with the robot motion interface. Unlike conventional mobile robot platforms, the use of a biped robot introduces additional challenges related to odometry quality, body motion, and sensor stability. The system was evaluated in an indoor environment using a LiDAR sensor and a ROS2-based navigation pipeline. The results demonstrate that classical navigation methods can be adapted to biped robots when proper frame transformations, localization, map representation, and command interfaces are considered.