A Unified Navigation and Obstacle Avoidance Framework for Humanoid Robots in Narrow Dynamic Environments
IEEE Transactions on Automation Science and Engineering · 2026 · 23 · 13419-13433
DOI: 10.1109/tase.2026.3714477 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Automation Science and Engineering, 2026. Verfasst von Jinye Hao und 4 weiteren Autor:innen. Die Arbeit nennt 50 Literaturangaben. Gefördert von: Science and Technology Innovation 2030—Project on Brain Science and Brain-Inspired Intelligence Research, Major Program of the National Natural Science Foundation of China, Excellent Youth Program of State Key Laboratory of Multimodal Artificial Intelligence Systems.
Abstract (Originalsprache)
Humanoid robots navigating in narrow, human-centered indoor environments must respond quickly when dynamic obstacles block the planned route, while maintaining accurate trajectory tracking under tight spatial constraints. This paper presents a unified navigation and obstacle avoidance framework for humanoid robots that integrates a multi-LiDAR SLAM system, a time-aware D* Lite planner, and a trajectory-aware reinforcement learning controller. The multi-LiDAR SLAM system effectively reduces occlusion-induced blind areas, enabling a more complete global map for reliable perception and planning. The time-aware D* Lite explicitly models the temporal behavior of obstacles, which enables low-latency online replanning and reduces path oscillations caused by fast-moving pedestrians. The tracking controller combines task-oriented observations and explicit reward shaping, allowing precise motion execution in narrow passages and large-turn trajectories. Experiments in a convenience store environment demonstrate robust obstacle avoidance and accurate tracking, achieving a trajectory tracking MAE of 0.0640 m and an average replanning time of 82.73 ms. The framework was further deployed in a real operating convenience store and achieved a 100% goal-reaching success rate across 20 trials. Note to Practitioners—This study addresses autonomous navigation for humanoid robots in narrow, human-centered indoor environments. The main practical difficulties are occlusion-induced perception blind areas, oscillatory planned trajectories caused by frequent replanning under dynamic blockages, and large trajectory tracking errors under bipedal locomotion constraints. We present an engineer-ready framework that combines three parts: a multi-LiDAR SLAM system that improves map completeness with a global 2.5D elevation grid map, a time-aware incremental planner that replans quickly while reducing path switching and oscillations, and a trajectory-aware RL controller that uses tracking-error feedback and task-oriented rewards to track paths smoothly in tight turns and narrow passages. Experiments show reliable avoidance of moving obstacles in store-like environments, and the system was deployed in a real operating convenience store, achieving a 100% goal-reaching success rate across 20 trials. Limitations include the need for accurate sensor calibration and time synchronization, and performance can degrade under persistent occlusions and highly unpredictable human motion. In future work, we will increase walking speed while preserving tracking stability and safety, and we will reduce planning latency to support faster and more frequent replanning.
Förderung
- Science and Technology Innovation 2030—Project on Brain Science and Brain-Inspired Intelligence Research
- Major Program of the National Natural Science Foundation of China
- Excellent Youth Program of State Key Laboratory of Multimodal Artificial Intelligence Systems
- National Natural Science Foundation of China
Zitieren
APA
Hao, J., Ji, S., Wang, Y., Wang, Y., & Qiao, H. (2026). A Unified Navigation and Obstacle Avoidance Framework for Humanoid Robots in Narrow Dynamic Environments. IEEE Transactions on Automation Science and Engineering, 23, 13419-13433. https://doi.org/10.1109/tase.2026.3714477
Vancouver
Hao J, Ji S, Wang Y, Wang Y, Qiao H. A Unified Navigation and Obstacle Avoidance Framework for Humanoid Robots in Narrow Dynamic Environments. IEEE Transactions on Automation Science and Engineering. 2026;23:13419-13433. doi:10.1109/tase.2026.3714477
BibTeX
@article{Hao2026,
author = {Hao, Jinye and Ji, Shenghao and Wang, Yifan and Wang, Yongxu and Qiao, Hong},
title = {A Unified Navigation and Obstacle Avoidance Framework for Humanoid Robots in Narrow Dynamic Environments},
journal = {IEEE Transactions on Automation Science and Engineering},
year = {2026},
volume = {23},
pages = {13419--13433},
doi = {10.1109/tase.2026.3714477}
}
EUVIMED