Learn to Quantify Social Interaction with Constraints for Pedestrian Walking

2026-06-16

AI Summary

Existing approaches to pedestrian trajectory prediction struggle to explicitly model and quantify diverse social interactions, limiting model robustness and interpretability. This work proposes a “Learn to Cluster” method that leverages unsupervised learning to automatically discover and cluster social interaction patterns directly from trajectory sequences. These patterns are represented as probabilistic latent variables and seamlessly integrated into an end-to-end trainable trajectory prediction framework. Notably, the approach requires no manual annotations, accommodates scenes with arbitrary numbers of agents, and is the first to enable automatic clustering and interpretation of social interactions. Experimental results demonstrate that the proposed method significantly improves long-term trajectory prediction accuracy across multiple benchmarks while enhancing the model’s capacity to represent and generalize complex social interactions.

Abstract

Long-term human path forecasting in crowds is critical for autonomous moving platforms (like autonomous driving cars and social robots) to avoid collision and make high-quality planning. Although the current research take into account social interactions for prediction, they don't reveal the exact kinds of social interactions happened among people and how the social interactions affect the decision-making process of pedestrians, which further limits its robustness. Social interactions in pedestrian walking are intuitively massive and hard to label and quantify. In this paper, we explore creatively to quantify and interpret how pedestrians interact with others by proposing Learn to Cluster. Our clustering social interactions is probabilistic latent variable generative, learning directly from sequential trajectory observations, scalable to arbitrary number of pedestrians. Learn to cluster is label-free and can be naturally integrated into the training process of the prediction model. The latent variables will then serve as 'labels' to categorize social interactions. Extensive experiments over several trajectory prediction benchmarks demonstrate that our method is able to learn the patterns of social interactions and effectively integrate the patterns to pedestrian trajectory prediction.

Problem

• social interaction
• pedestrian trajectory prediction
• human path forecasting
• crowd behavior
• interaction quantification

Innovation

• Learn to Cluster
• social interaction quantification
• latent variable modeling
• label-free learning
• pedestrian trajectory prediction