Learning to Evolve: Multi-modal Interactive Fields for Robust Humanoid Navigation in Dynamic Environments

2026-05-20

AI Summary

This work addresses the challenge of unreliable scene memory in humanoid robot navigation within dynamic environments, where gait-induced disturbances, environmental changes, and interaction safety constraints degrade localization and mapping performance. To overcome this, the authors propose a Multimodal Interaction Field (MIF) architecture that integrates confidence-aware semantic 3D Gaussian splatting, a topology-preserving spatial field, and an interaction-safety-aware geometric field. A discrepancy detection mechanism is introduced to effectively distinguish gait-induced artifacts from genuine environmental changes, enabling efficient updates of locally inconsistent regions. Evaluated in real-world dynamic office settings, the system improves relocalization success rate from 12% to 94% while reducing semantic memory consumption by 91.4%, substantially enhancing navigation robustness and online computational efficiency in dynamic scenarios.

Abstract

Safe manipulation-oriented navigation for humanoid robots requires scene memory that remains reliable under locomotion-induced perceptual distortion, environmental changes, and interaction-level geometric safety constraints. Existing semantic mapping and scene-graph systems are difficult to deploy directly in this setting because they often assume stable camera trajectories, static environments, or coarse object geometry. We introduce the Multi-modal Interactive Field (MIF), a humanoid-oriented system that integrates confidence-aware semantic 3D Gaussian Splatting, discrepancy-triggered spatial memory updates, and task-driven geometric reconstruction within a closed-loop perception-adaptation pipeline. MIF couples three fields: an uncertainty-aware 3DGS Appearance Field that suppresses gait-induced blur, a Spatial Field that maintains topological memory, and a Geometry Field that supports Interaction Pose Safety (IPS) before manipulation. A discrepancy detection score is introduced to separate locomotion-induced false-positive changes from persistent changes and updates only locally inconsistent regions. On a Unitree-G1 humanoid in a real dynamic office, MIF improves relocation success in non-static environments from 12% to 94% compared with static scene-graph memory, while reducing semantic memory footprint by 91.4% through feature distillation for practical online operation. Project page and code: https://ziya-jiang.github.io/MIF-homepage/

Problem

• humanoid navigation
• dynamic environments
• scene memory
• perceptual distortion
• geometric safety constraints

Innovation

• Multi-modal Interactive Field
• 3D Gaussian Splatting
• discrepancy detection
• Interaction Pose Safety
• semantic memory distillation