Technology Category
Humans and AI: Human-Aware Planning and Behavior PredictionPlanning, Routing, and Scheduling: Model-Based ReasoningSearch and Optimization: Learning to Search
ShibuyaSocial: Multi-scale Model of Pedestrian Flows in Scramble Crossing
🤖 AI Summary
Existing pedestrian flow modeling for Shibuya’s omnidirectional intersection suffers from a disconnection between global path planning and local collision avoidance, and fails to accurately capture traffic-signal–driven start-stop dynamics. Method: We propose the first attention-driven, multi-scale behavioral learning model for pedestrian flow, unifying route choice, local interaction, and signal-state transitions within a single framework. Our approach integrates spatiotemporal graph neural networks with a multi-scale feature alignment mechanism, and—novelty—we employ attention to jointly predict global trajectories and local avoidance maneuvers. Contribution/Results: Evaluated on real-world Shibuya video-based trajectory data, our model significantly outperforms single-scale baselines, achieving state-of-the-art performance: 23.6% reduction in average displacement error (ADE) and 19.8% reduction in final displacement error (FDE), alongside superior fidelity in simulating signal-responsive start-stop dynamics. The model establishes an interpretable, scalable paradigm for intelligent management of high-density urban pedestrian flows.
Technology Category
Application Category
📝 Abstract
This paper presents a learning-based model of pedestrian flows that integrates multi scale behaviors such as global route selection and local collision avoidance in urban spaces, particularly focusing on pedestrian movements at Shibuya scramble crossing. Since too much congestion of pedestrian flows can cause serious accidents, mathematically modeling and predicting pedestrian behaviors is important for preventing such accidents and providing a safe and comfortable environment. Although numerous studies have investigated learning-based modeling methods, most of them focus only on the local behavior of pedestrians, such as collision avoidance with neighbors and environmental objects. In an actual environment, pedestrian behavior involves more complicated decision making including global route selection. Moreover, a state transition from stopping to walking at a traffic light should be considered simultaneously. In this study, the proposed model integrates local behaviors with global route selection, using an Attention mechanism to ensure consistent global and local behavior predictions. We recorded video data of pedestrians at Shibuya scramble crossing and trained the proposed model using pedestrian walking trajectory data obtained from the video. Simulations of pedestrian behaviors based on the trained model qualitatively and quantitatively validated that the proposed model can appropriately predict pedestrian behaviors.
Problem
Research questions and friction points this paper is trying to address.
Model pedestrian flows integrating global route selection and local collision avoidance
Predict pedestrian behaviors to prevent accidents in crowded urban crossings
Address the gap between local and global behavior modeling in pedestrian simulations
Innovation
Methods, ideas, or system contributions that make the work stand out.
Integrates global route selection with local collision avoidance
Uses Attention mechanism for consistent behavior predictions
Trained on video data from Shibuya scramble crossing
🔎 Similar Papers
No similar papers found.
