Human-Robot Cooperative Distribution Coupling for Hamiltonian-Constrained Social Navigation

TL;DR

This work tackles the challenge of socially aware navigation in human crowds by introducing NaviDIFF, a framework that fuses port-Hamiltonian physics with diffusion-based modeling of human-robot cooperation and a spatial-temporal transformer to capture social-temporal dynamics. A diffusion network models cooperative uncertainty while a Hamiltonian-constrained controller provides energy-balanced, stable robot behavior, with RLHF refining policies to align with human preferences. The approach offers interpretable energy-based dynamics, closed-loop stability, and improved performance over state-of-the-art baselines in both simulation and real-world tests. The integration of an LLM-driven human-robot interface and diffusion-guided policy synthesis enables adaptable, norm-compliant navigation with practical implications for deployment in crowded environments.

Abstract

Navigating in human-filled public spaces is a critical challenge for deploying autonomous robots in real-world environments. This paper introduces NaviDIFF, a novel Hamiltonian-constrained socially-aware navigation framework designed to address the complexities of human-robot interaction and socially-aware path planning. NaviDIFF integrates a port-Hamiltonian framework to model dynamic physical interactions and a diffusion model to manage uncertainty in human-robot cooperation. The framework leverages a spatial-temporal transformer to capture social and temporal dependencies, enabling more accurate spatial-temporal environmental dynamics understanding and port-Hamiltonian physical interactive process construction. Additionally, reinforcement learning from human feedback is employed to fine-tune robot policies, ensuring adaptation to human preferences and social norms. Extensive experiments demonstrate that NaviDIFF outperforms state-of-the-art methods in social navigation tasks, offering improved stability, efficiency, and adaptability.

Figures (5)

• Figure 1: An illustration of a social navigation task: the mobile robot must navigate through pedestrians while maintaining appropriate social distances and finding a feasible path.
• Figure 2: NaviDIFF Architecture: NaviDIFF leverages a spatial-temporal transformer to represent Hamiltonian terms, capturing HRI environmental dynamics and ensuring closed-loop stability. It also addresses cooperative uncertainty using a diffusion neural network. Additionally, the framework incorporates reinforcement learning from human feedback (RLHF) and utilizes a large language model (LLM) for enhanced performance.
• Figure 3: Experiment Results: Success rate (green), collision rate (red), timeout rate (yellow), and social score (orange) for each policy, based on 500 test runs under the same environment configuration.
• Figure 4: Comparison of Trajectories Visualization: Visualization of the trajectories for NaviDIFF and two ablation models, tested on two different scenarios. Zoom in for better readability.
• Figure 5: Configuration of the physical mobile robot platform.