Great GATsBi: Hybrid, Multimodal, Trajectory Forecasting for Bicycles using Anticipation Mechanism

2025-08-20

AI Summary

Bicycle trajectory prediction remains underexplored despite growing demands for non-motorized vehicle safety and intelligent perception. Method: This paper proposes a dual-modal prediction framework integrating physical dynamics and social interaction. It introduces a novel future-intent anticipation mechanism grounded in psychological and sociological principles, jointly modeling attenuated historical states and anticipated trajectories of neighboring vehicles via a graph attention network (GAT). The framework synergistically combines a physics-based dynamical model with a data-driven social interaction model, and employs a multi-modal prediction architecture to balance short-term motion accuracy and long-term behavioral consistency. Contribution/Results: Evaluated on a real-world cycling dataset, our method outperforms state-of-the-art approaches: short-term prediction error decreases by 18.3%, and long-term trajectory plausibility improves by 24.7%. The framework delivers interpretable, robust predictions, establishing a new paradigm for intelligent non-motorized vehicle perception and traffic safety analysis.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems

Abstract

Accurate prediction of road user movement is increasingly required by many applications ranging from advanced driver assistance systems to autonomous driving, and especially crucial for road safety. Even though most traffic accident fatalities account to bicycles, they have received little attention, as previous work focused mainly on pedestrians and motorized vehicles. In this work, we present the Great GATsBi, a domain-knowledge-based, hybrid, multimodal trajectory prediction framework for bicycles. The model incorporates both physics-based modeling (inspired by motorized vehicles) and social-based modeling (inspired by pedestrian movements) to explicitly account for the dual nature of bicycle movement. The social interactions are modeled with a graph attention network, and include decayed historical, but also anticipated, future trajectory data of a bicycles neighborhood, following recent insights from psychological and social studies. The results indicate that the proposed ensemble of physics models -- performing well in the short-term predictions -- and social models -- performing well in the long-term predictions -- exceeds state-of-the-art performance. We also conducted a controlled mass-cycling experiment to demonstrate the framework's performance when forecasting bicycle trajectories and modeling social interactions with road users.

Problem

• Predicting bicycle trajectories using hybrid multimodal models
• Addressing bicycle movement's dual nature via physics and social modeling
• Improving road safety by forecasting bicycle interactions with surroundings

Innovation

• Hybrid physics and social modeling
• Graph attention network interactions
• Anticipated future trajectory data