A field approach for pedestrian movement modelling

TL;DR

This paper introduces a novel physics-based field theory approach for modeling pedestrian movement, offering a flexible and general framework that can be integrated with data assimilation and neural networks for improved crowd analysis.

Contribution

The paper presents a new field theory-based method for pedestrian movement modeling, deriving differential equations and exploring neural network integration for enhanced analysis.

Findings

Formulated differential equations for pedestrian movement using field theory.

Demonstrated the potential for data assimilation with analytical transition functions.

Discussed neural network integration for flexible and extendable modeling.

Abstract

There are different physics-based approaches for analysing pedestrian movement. Physics-based methods like statistical mechanics-based models apply the laws of physics to drive equations for analysing crowd behaviour. This paper will introduce a physics-based approach based on field theory as a new tool for crowd analysis to determine governing differential equations. Formulating the pedestrian movement with differential equations has a primary advantage for data assimilation techniques because some of these methods only work with models with analytical transition functions, which are obtained by incorporating a field approach. Furthermore, the field approach provides more generality since the field could be any scalar field. Several Lagrangians are presented in this work, and the primary purpose was to lay the groundwork for this new type of thinking. Furthermore, as pedestrian movement is mainly unregulated, the approach presented in the paper can be valuable for future development since the Lagrangian could be explicitly obtained for pedestrian movement spaces such as train stations and shopping malls. Finally, we discuss a general approach for predicting the action and how neural networks might play a role, which brings more flexibility and extendibility to our approach.