HiCrowd: Hierarchical Crowd Flow Alignment for Dense Human Environments
Yufei Zhu, Shih-Min Yang, Martin Magnusson, Allan Wang
AI summary
Problem
Mobile robots face the freezing robot problem in dense crowds, where reactive planners and flat learning methods struggle to capture complex human interactions or guarantee safety. This leaves robots stuck or inefficient when navigating unstructured, high-density pedestrian environments.
Approach
HiCrowd uses a hierarchical framework where a high-level reinforcement learning policy predicts a follow point to align with compatible pedestrian groups, while a low-level model predictive control module safely tracks this guidance over a short horizon.
Key results
- Achieves 100% success rate with zero freezing in offline and online simulations
- Reduces navigation time and path length by up to 30% compared to baselines
- Successfully deployed in real-world crowded environments without retraining
- Demonstrates that crowd-following rewards accelerate learning and improve generalization
Why it matters
Provides a scalable, socially aware navigation principle for service and delivery robots operating in unstructured, high-density public spaces.
Abstract
Navigating through dense human crowds remains a significant challenge for mobile robots. A key issue is the freezing robot problem, where the robot struggles to find safe motions and becomes stuck within the crowd. To address this, we propose HiCrowd, a hierarchical framework that integrates reinforcement learning (RL) with model predictive control (MPC). HiCrowd leverages surrounding pedestrian motion as guidance, enabling the robot to align with compatible crowd flows. A high-level RL policy generates a follow point to align the robot with a suitable pedestrian group, while a low-level MPC safely tracks this guidance with short horizon planning. The method combines long-term crowd aware decision making with safe short-term execution. We evaluate HiCrowd against reactive and learning-based baselines in offline setting (replay- ing recorded human trajectories) and online setting (human trajectories are updated to react to the robot in simulation). Ex- periments on a real-world dataset and a synthetic crowd dataset show that our method outperforms in navigation efficiency and safety, while reducing freezing behaviors. We further validate through real-world deployment in a public museum and Expo 2025 Osaka, where it navigates dense pedestrian flows without retraining, demonstrating robust and socially aware behavior. Our results suggest that leveraging human motion as guidance, rather than treating humans solely as dynamic obstacles, provides a powerful principle for safe and efficient robot navigation in crowds. Project code and demos are available at https://github.com/test-bai-cpu/HiCrowd.