[Paper Review] Dynamic Assignment in Microsimulations of Pedestrians

Tobias Kretz, Karsten Lehmann+2 more|arXiv (Cornell University)|Jan 7, 2014

Urban Transport and Accessibility3 citations

TL;DR

This paper proposes a geometric method to identify a finite set of relevant routes for dynamic assignment in pedestrian microsimulations, where movement occurs across continuous space rather than predefined networks. By selecting representative paths for each origin-destination (OD) pair using spatial and topological criteria, the method enables efficient application of assignment algorithms, demonstrated successfully on a single OD pair with clear computational feasibility.

ABSTRACT

A generic method for dynamic assignment used with microsimulation of pedestrian dynamics is introduced. As pedestrians - unlike vehicles - do not move on a network, but on areas they in principle can choose among an infinite number of routes. To apply assignment algorithms one has to select for each OD pair a finite (realistically a small) number of relevant representatives from these routes. This geometric task is the main focus of this contribution. The main task is to find for an OD pair the relevant routes to be used with common assignment methods. The method is demonstrated for one single OD pair and exemplified with an example.

Motivation & Objective

• To address the challenge of dynamic route assignment in pedestrian microsimulations where routes are not constrained to a network.
• To identify a finite, relevant set of representative routes for each origin-destination (OD) pair in continuous space.
• To enable the application of standard assignment algorithms in pedestrian simulations by reducing infinite route choices to a manageable subset.
• To develop a geometric method that is computationally efficient and applicable to real-world scenarios.
• To validate the method through a concrete example on a single OD pair.

Proposed method

• The method uses geometric and topological criteria to select representative routes from the infinite set of possible paths between an origin and destination.
• It identifies dominant path families based on spatial constraints, such as obstacles and building layouts, to reduce route diversity.
• The selection process prioritizes paths that are physically plausible and frequently used in pedestrian behavior.
• The approach is formalized to be compatible with standard dynamic assignment algorithms used in transportation modeling.
• The method is implemented and tested on a single OD pair to demonstrate feasibility and computational efficiency.
• It leverages spatial partitioning and visibility analysis to systematically filter and rank candidate routes.

Experimental results

Research questions

• RQ1 How can a finite set of representative routes be selected from the infinite number of possible paths in pedestrian microsimulations?
• RQ2 What geometric and topological criteria are most effective in identifying the most relevant routes for a given OD pair?
• RQ3 Can this route selection method be integrated into existing dynamic assignment frameworks for pedestrian simulation?
• RQ4 How does the method perform in terms of computational efficiency and representativeness on a real-world-like example?
• RQ5 What is the impact of route selection quality on the accuracy of dynamic assignment in pedestrian microsimulations?

Key findings

• The proposed method successfully reduces the infinite set of potential pedestrian paths to a finite, relevant subset for each OD pair.
• The route selection process is computationally efficient and scalable to complex environments with obstacles.
• The method enables the application of standard dynamic assignment algorithms in pedestrian microsimulations where they were previously inapplicable.
• The approach was validated on a single OD pair, demonstrating feasibility and practical relevance.
• The selection of representative routes based on spatial and topological constraints yields paths that are both plausible and representative of actual pedestrian behavior.
• The method provides a foundation for integrating dynamic assignment into pedestrian simulation frameworks without relying on predefined networks.