Following Is All You Need: Robot Crowd Navigation Using People As Planners
Yuwen Liao, Xinhang Xu, Ruofei Bai, Yizhuo Yang, Muqing Cao, Shenghai Yuan, Lihua Xie
AI summary
Problem
Existing crowd navigation methods rely on computationally heavy predictive models or reinforcement learning that struggle in dense, dynamic settings, while overlooking the readily available intelligence of nearby pedestrians.
Approach
The system continuously evaluates nearby pedestrians using rule-based criteria to identify an optimal leader, then uses a lightweight local planner to follow that leader through short-horizon subgoals.
Key results
- Achieves lowest collision counts and highest efficiency across simulated crowded scenes compared to five baselines
- Maintains safe distances from heterogeneous road users without explicit type-aware modeling
- Generates socially compliant, human-like navigation behaviors without hardcoding social norms
- Validated through both simulation and real-world experiments
Why it matters
Provides a computationally lightweight and robust alternative for safe robot deployment in dense public spaces where heavy AI planners are impractical.
Abstract
Navigating in crowded environments requires the robot to be equipped with high-level reasoning and planning tech- niques. Existing works focus on developing complex and heavy- weight planners while ignoring the role of human intelligence. Since humans are highly capable agents who are also widely available in a crowd navigation setting, we propose an alternative scheme where the robot utilises people as planners to benefit from their effective planning decisions and social behaviours. Through a set of rule-based evaluations, we identify suitable human leaders who exhibit the potential to guide the robot towards its goal. Using a simple base planner, the robot follows the selected leader through short-horizon subgoals that are designed to be straightforward to achieve. We demonstrate through both simulated and real-world experiments that our novel framework generates safe and efficient robot plans compared to existing planners, even without predic- tive or data-driven modules. Our method also brings human-like robot behaviours without explicitly defining traffic rules and social norms.