Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PolygMap: A Perceptive Locomotion Framework for Humanoid Robot Stair Climbing

Bingquan Li, Ning Wang, Tianwei Zhang, Zhicheng He, Yucong Wu

TL;DR

The paper addresses the challenge of humanoid stair climbing by introducing a perception-based locomotion framework that fuses LiDAR and RGB-D data to construct a polygonal staircase plane semantic map, from which footholds are generated and integrated into real-time footstep planning. It combines multi-sensor fusion–based state estimation, plane-based foothold extraction via RANSAC, and a gait planner that generates smooth torso-foot trajectories at 20-30 Hz on an onboard NVIDIA Orin. Major contributions include the polygonal stair map construction, erosion-based foothold extraction, and a real-time foothold-aware trajectory planner, validated through Gazebo simulations and real-world indoor/outdoor stair climbs. The framework demonstrates robust stair climbing under perception noise and varying conditions, enabling reliable operation on long staircases with minimal footstep errors. Open-source release is planned to facilitate adoption and further research in perception-guided humanoid locomotion.

Abstract

Recently, biped robot walking technology has been significantly developed, mainly in the context of a bland walking scheme. To emulate human walking, robots need to step on the positions they see in unknown spaces accurately. In this paper, we present PolyMap, a perception-based locomotion planning framework for humanoid robots to climb stairs. Our core idea is to build a real-time polygonal staircase plane semantic map, followed by a footstep planar using these polygonal plane segments. These plane segmentation and visual odometry are done by multi-sensor fusion(LiDAR, RGB-D camera and IMUs). The proposed framework is deployed on a NVIDIA Orin, which performs 20-30 Hz whole-body motion planning output. Both indoor and outdoor real-scene experiments indicate that our method is efficient and robust for humanoid robot stair climbing.

PolygMap: A Perceptive Locomotion Framework for Humanoid Robot Stair Climbing

TL;DR

The paper addresses the challenge of humanoid stair climbing by introducing a perception-based locomotion framework that fuses LiDAR and RGB-D data to construct a polygonal staircase plane semantic map, from which footholds are generated and integrated into real-time footstep planning. It combines multi-sensor fusion–based state estimation, plane-based foothold extraction via RANSAC, and a gait planner that generates smooth torso-foot trajectories at 20-30 Hz on an onboard NVIDIA Orin. Major contributions include the polygonal stair map construction, erosion-based foothold extraction, and a real-time foothold-aware trajectory planner, validated through Gazebo simulations and real-world indoor/outdoor stair climbs. The framework demonstrates robust stair climbing under perception noise and varying conditions, enabling reliable operation on long staircases with minimal footstep errors. Open-source release is planned to facilitate adoption and further research in perception-guided humanoid locomotion.

Abstract

Recently, biped robot walking technology has been significantly developed, mainly in the context of a bland walking scheme. To emulate human walking, robots need to step on the positions they see in unknown spaces accurately. In this paper, we present PolyMap, a perception-based locomotion planning framework for humanoid robots to climb stairs. Our core idea is to build a real-time polygonal staircase plane semantic map, followed by a footstep planar using these polygonal plane segments. These plane segmentation and visual odometry are done by multi-sensor fusion(LiDAR, RGB-D camera and IMUs). The proposed framework is deployed on a NVIDIA Orin, which performs 20-30 Hz whole-body motion planning output. Both indoor and outdoor real-scene experiments indicate that our method is efficient and robust for humanoid robot stair climbing.
Paper Structure (13 sections, 36 equations, 8 figures, 2 tables)

This paper contains 13 sections, 36 equations, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Plane Semantic Mapping
  4. Perspective Locomotion
  5. Methods
  6. Multi-Sensor Fusion based State Estimation
  7. Foothold Generation
  8. Footstep Planning
  9. Experiments Results and Evaluations
  10. Sensor Fusion Experiment Analysis
  11. Gazebo Simulation Results
  12. Real World Stair Climbing Evaluations
  13. Conclusions

Figures (8)

  • Figure 1: Perceptive Locomotion for Humanoid Stair Climbing
  • Figure 2: The system integrates joint recorders, depth sensing and LIO estimator. Robot pose is obtained via fusing forward kinematics and the LIO, while depth images provide a polygonal map for foothold generation. Based on these footholds, footstep and whole body motion planning enable stable stair climbing.
  • Figure 3: Polygmap-based footstep motion planning logic
  • Figure 4: Comparative Analysis of Position Estimation: Ontology-Based, LIO, and the proposed Fusion Method
  • Figure 5: Simulation experiment of the proposed system. The figure shows the perception--planning--control pipeline and stair climbing locomotion sequence.
  • ...and 3 more figures