Real-Time Navigation and Mapping for Humanoid Robots: A Case Study Using RTAB-Map on SURENA V

Advancements in humanoid robotics have significantly improved mobility, perception, and interaction, bringing these systems closer to practical, real-world applications. Among these capabilities, navigation remains a cornerstone, enabling robots to autonomously interact with dynamic and unstructured environments. Achieving reliable navigation in humanoid robots is particularly challenging due to the ever-changing nature of real-world settings. Traditional ۲D SLAM techniques often fail to provide sufficient adaptability and precision in such scenarios, necessitating the adoption of more robust and versatile methods. RTAB-Map, a graph-based SLAM algorithm, is designed to address these challenges by constructing and maintaining a ۳D map of the environment while performing real-time localization. Its ability to integrate visual and depth data ensures robust performance in complex, dynamic environments, making it ideal for humanoid robot navigation. In this study, the SURENA humanoid robot, equipped with an RGB-D camera, is employed to validate the proposed framework. The robot is modelled using the Linear Inverted Pendulum Model (LIPM), and footstep trajectories are generated using the Divergent Component of Motion (DCM) algorithm. Simultaneous localization and mapping are performed using RTAB-Map to generate both a ۲D occupancy grid and a ۳D Octomap. Once the target destination is specified, global paths are planned using Dijkstra and A* algorithms, while the Dynamic Window Approach (DWA) handles local motion planning. The entire framework is implemented and validated in the ROS and Choreonoid environments, demonstrating the feasibility of achieving robust, online navigation for humanoid robots.

کلیدواژه ها:

humanoid robot ، RTAB-Map ، SLAM ، online navigation ، path planning