A unified pedestrian routing model for graph-based wayfinding built on cognitive principles

ABSTRACT

The wayfinding behavior of pedestrians in street and building networks can be predicted by computer simulations based on routing models. To model realistic routing behavior, it is necessary to integrate spatial- and social-cognitive aspects into the wayfinding models. However, a model that incorporates diverse influencing factors on pedestrian route planning has yet not been developed for microscopic simulations. We present a unified routing model that describes pedestrian route choices in street and building environments by integrating spatial- and social-cognitive aspects. We achieve an integration of both domains by combining different graph-based routing methods, each formalizing a cognitive theory. In addition, we present a calibration method for the spatial-cognitive aspects. For validation purposes, we use the model to simulate how the visitors of a music festival navigate to the event and how people navigate in a city district. Our methodology is highly flexible and can be extended to include other aspects of wayfinding behavior.

KEYWORDS:

• Pedestrian behavior modeling
• pedestrian behavior simulation
• cognitive modeling
• model integration
• graph-based wayfinding

Acknowledgments

We would like to thank our student assistants for helping with the data acquisition during the festival. In addition, we want to thank all participants of the MultikOSi research project for their discussions and support. We especially thank Prof. Annette Spellerberg and her research team for sharing their questionnaire data. Furthermore, we like to thank Prof. Hölscher and his research team for fruitful discussions regarding cognition in human wayfinding. Also, we thank our student assistants, who contributed to the pedestrian simulation framework ‘MomenTUMv2’. Finally, we would like to thank the anonymous reviewers for their helpful and high-quality suggestions and comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. We removed the error term because it creates non-deterministic noise to the model and tends to obstruct proper validation. We are investigating this error term and our current hypothesis is that this is strongly related to the angular and distance error in human perception.

2. There are several other approaches for clustering the routes, e.g. translating the routing paths pairwise into polygons and checking if obstacles reside within the combined polygon.

3. In some cases, the factor combinations emphasize lower-knowledge routing behavior in such a way that simulated pedestrians will get lost. Because we did not integrate a spatial learning concept, these lost pedestrians cannot recover. We will address this topic in further research.

4. Using the highest survey-based factor combination will be beneficial for integrating the social-cognitive aspects later on.