COLSON: Controllable Learning-Based Social Navigation via Diffusion-Based Reinforcement Learning

2025-03-18

AI Summary

Existing deep reinforcement learning (DRL) approaches for socially aware navigation of mobile robots in dynamic pedestrian environments suffer from limited action flexibility due to restrictive Gaussian distribution assumptions. To address this, we propose the first diffusion-model-based framework for social navigation RL—replacing fixed-distribution priors with a flexible, controllable generative process for continuous action synthesis and real-time adaptation. Our method enables post-training action smoothing optimization and zero-shot transfer to static obstacle scenarios, significantly enhancing generalization and robustness. Experiments demonstrate a 12.7% improvement in navigation success rate over Gaussian-policy baselines in dense dynamic environments, while maintaining robust performance on static obstacles without fine-tuning.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Reinforcement Learning

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Mobile robot navigation in dynamic environments with pedestrian traffic is a key challenge in the development of autonomous mobile service robots. Recently, deep reinforcement learning-based methods have been actively studied and have outperformed traditional rule-based approaches owing to their optimization capabilities. Among these, methods that assume a continuous action space typically rely on a Gaussian distribution assumption, which limits the flexibility of generated actions. Meanwhile, the application of diffusion models to reinforcement learning has advanced, allowing for more flexible action distributions compared with Gaussian distribution-based approaches. In this study, we applied a diffusion-based reinforcement learning approach to social navigation and validated its effectiveness. Furthermore, by leveraging the characteristics of diffusion models, we propose an extension that enables post-training action smoothing and adaptation to static obstacle scenarios not considered during the training steps.

Problem

• Mobile robot navigation in dynamic pedestrian environments.
• Limitations of Gaussian distribution in continuous action spaces.
• Enhancing flexibility and adaptability in social navigation tasks.

Innovation

• Diffusion-based reinforcement learning for social navigation
• Post-training action smoothing using diffusion models
• Adaptation to static obstacles post-training