An extensible simulation environment and movement metrics for testing walking behavior in agent-based models

https://doi.org/10.1016/j.compenvurbsys.2011.07.005Get rights and content

Abstract

Human movement is a significant ingredient of many social, environmental, and technical systems, yet the importance of movement is often discounted in considering systems’ complexity. Movement is commonly abstracted in agent-based modeling (which is perhaps the methodological vehicle for modeling complex systems), despite the influence of movement upon information exchange and adaptation in a system. In particular, agent-based models of urban pedestrians often treat movement in proxy form at the expense of faithfully treating movement behavior with realistic agency. There exists little consensus about which method is appropriate for representing movement in agent-based schemes. In this paper, we examine popularly-used methods to drive movement in agent-based models, first by introducing a methodology that can flexibly handle many representations of movement at many different scales and second, introducing a suite of tools to benchmark agent movement between models and against real-world trajectory data. We find that most popular movement schemes do a relatively poor job of representing movement, but that some schemes may well be “good enough” for some applications. We also discuss potential avenues for improving the representation of movement in agent-based frameworks.

Highlights

► We present a novel framework for simulating pedestrians and metrics for evaluating movement. ► Our approach can be applied across application scenarios, cities, and scales. ► We prove its usefulness in studying a range of movement scenarios at different scales.

Introduction

Movement of pedestrians is significant across a variety of domains in which it is infeasible to experiment with real people or environments. As an alternative, agent-based models (Russell & Norvig, 1995), which date to Alan Turing’s original work on intelligent machines (Turing, 1950), are popularly used to generate synthetic pedestrians in simulation. In many instances, however, representation of agent movement in models is cursory compared to our understanding of the factors that drive human motion in the real world and little robust investigation of the plausibility of pedestrian movement in agent-based models has been performed. Rather than deriving from behavior, pedestrian agent-based models are often developed from the physics or informatics of movement. The reasoning for this is straightforward: motion is well-understood in these domains. However, there exists little basis for developing consensus among the builders of agent-based models regarding the implications of choosing one movement algorithm over another. It is also troublesome, philosophically, when models of human movement bear little behavioral resemblance to reality, but are used to inform decisions.

The question that we pose in this paper is: which algorithms are appropriate proxies for human movement in agent-based models and why? With this in mind, we critically examine movement algorithms commonly employed in agent-based pedestrian models, assessing their fit with theory, with each other, and—using traces of real human movement—with reality. Answering this question first requires that we develop an extensible agent-based modeling platform that can realize multiple models, algorithms, and parameters. Second, it requires that we introduce methods to assess the relative performance of movement algorithms and their ability to replicate real-world paths. We will demonstrate that, with few exceptions, movement algorithms for agent-based models do a relatively poor job of reproducing realistic mobility in simulation. Some algorithms are perhaps “good enough” as rough proxies for human movement, but this generally holds only for particular types of sub-movement, at particular scales, or in specific environments. We examine why this might be the case and what might be done to remedy the problem.

Section snippets

Related work

A variety of approaches have been developed to handle pedestrian movement in agent-based simulations. Physics models often work from the assumption that agents “go with the flow” of an ambient crowd and that their motion can be modeled using equations for flow of non-human media such as gases or fluids (which are more tractable than the “messy complexity” of people). A related assumption is often made: that pedestrians might cede their personal movement behavior to that of the crowd in very

Methods

In this section, we will introduce an extensible automata scheme for modeling movement. We will also detail how we fit common algorithms to this framework.

Results

In order to investigate the relative performance of the movement algorithms described in the previous section, we first considered their correspondence with theory and the plausibility of their movement representations. Second, we compared them empirically and substantively to each other. Third, we examined the extent to which they might match recorded movements of people in the real-world. Because different movement routines are designed to operate at different scales, we will discuss the

Conclusions

In this paper, we examined popular algorithms as motion controllers for synthetic pedestrians in agent-based models, with the intention of assessing whether they faithfully represent real-world movement. This required development of an extensible agent-based modeling platform that could represent varying movement schemes, as well as a suite of analytical tools that could relate simulated movement between models and relative to real-world pedestrian trajectories.

The modeling scheme that we

Acknowledgements

This material is based in part on work supported by the National Science Foundation under grant numbers 1002517, 0643322, and 0624208. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

References (102)

  • F. Bartumeus et al.

    Animal search strategies: A quantitative random-walk analysis

    Ecology

    (2005)
  • M. Batty

    Agent-based pedestrian modelling

  • M. Batty et al.

    The discrete dynamics of small-scale spatial events: Agent-based models of mobility in carnivals and street parades

    International Journal of Geographical Information Science

    (2003)
  • M. Batty et al.

    Safety in numbers? Modelling crowds and designing control for the Notting Hill carnival

    Urban Studies

    (2003)
  • J. Bohannan

    Directing the herd: crowds and the science of evacuation

    Science

    (2005)
  • Boulic, R., Noser, H., Thalmann, D. (1994). Automatic derivation of curved human walking trajectories from synthetic...
  • D. Brockman et al.

    The scaling laws of human travel

    Nature

    (2006)
  • Chenney, S. (2004). Flow tiles. In: Proceedings of the 2004 ACM SIGGRAPH/Eurographics symposium on Computer Animation,...
  • T.M. Ciolek

    Pedestrian behaviour in pedestrian spaces: some findings of a naturalistic field study

  • W.A.V. Clark et al.

    Understanding the social context of the Schelling segregation model

    Proceedings of the National Academy of Sciences

    (2008)
  • C. Couch

    Collective behavior: An examination of some stereotypes

    Social Problems

    (1968)
  • A. Crooks et al.

    Agent street: An environment for exploring agent-based models in second life

    Journal of Artificial Societies and Social Simulation

    (2009)
  • J.E. Cutting et al.

    How we avoid collisions with stationary and moving obstacles

    Psychological Review

    (1995)
  • J.M. Dabbs et al.

    Beauty is power: the use of space on the sidewalk

    Sociometry

    (1975)
  • E.W. Dijkstra

    A note on two problems in connection with graphs

    Numerische Mathematik

    (1959)
  • R.M. Eggo et al.

    Spatial dynamics of the 1918 influenza pandemic in England, Wales and the United States

    Journal of The Royal Society Interface

    (2011)
  • J.M. Epstein

    Modeling civil violence. An agent-based computational approach

    Proceedings of the National Academy of Science

    (2002)
  • J.M. Epstein et al.

    Coupled contagion dynamics of fear and disease: Mathematical and computational explorations

    PLoS ONE

    (2008)
  • I. Farkas et al.

    Crowd behaves as excitable media during Mexican wave

    Nature

    (2002)
  • N.M. Ferguson et al.

    Strategies for mitigating an influenza pandemic

    Nature

    (2006)
  • J.J. Fruin

    Pedestrian planning and design

    (1971)
  • T. Gärling et al.

    Distance minimization in downtown pedestrian shopping

    Environment and Planning A

    (1988)
  • E. Goffmann

    Relations in public: Microstudies in the public order

    (1971)
  • R.A. Golledge

    Wayfinding behavior: Cognitive mapping and other spatial processes

    (1999)
  • M.C. González et al.

    Understanding individual human mobility patterns

    Nature

    (2008)
  • D.M. Gorman et al.

    Agent-based modeling of drinking behavior: A preliminary model and potential applications to theory and practice

    American Journal of Public Health

    (2006)
  • B.T. Grenfell et al.

    Travelling waves and spatial hierarchies in measles epidemics

    Nature

    (2001)
  • S.J. Guy et al.

    “PLEdestrians: A least-effort approach to crowd simulation

  • M. Haklay et al.

    “So go downtown”: Simulating pedestrian movement in town centres

    Environment and Planning B

    (2001)
  • P.E. Hart et al.

    A formal basis for the heuristic determination of minimum cost paths

    IEEE Transactions on Systems, Science and Cybernetics

    (1968)
  • D. Helbing

    A fluid-dynamic model for the movement of pedestrians

    Complex Systems

    (1992)
  • D. Helbing et al.

    Social force model for pedestrian dynamics

    Physical Review E

    (1995)
  • D. Helbing et al.

    Self-organization phenomena in pedestrian crowds

  • D. Helbing et al.

    Self-organizing pedestrian movement

    Environment and Planning B

    (2001)
  • L.F. Henderson

    The statistics of crowd fluids

    Nature

    (1971)
  • G. Hess

    Disease in metapopulation models: implications for conservation

    Ecology

    (1996)
  • B. Hillier et al.

    The social logic of space

    (1984)
  • S.P. Hoogendoorn et al.

    Gas-kinetic modeling and simulation of pedestrian flows

    Transportation Research Record

    (2000)
  • S.P. Hoogendoorn et al.

    Edestrian behavior at bottlenecks

    Transportation Science

    (2005)
  • R.L. Hughes

    The flow of human crowds

    Annual Review of Fluid Mechanics

    (2003)
  • Cited by (62)

    • Inverse augmentation: Transposing real people into pedestrian models

      2023, Computers, Environment and Urban Systems
      Citation Excerpt :

      We also calculated autocorrelation of their sinuosity, using (8) mean cosine of turning angle between successive steps, (9) probability of turning in the same direction per step, and (9) correlation among the cosines of successive turn angles (Caldwell & Nams, 2006; Nams, 2006a). For comparative validation, we assessed the performance of user participants in the simulation against different real-world walking and urban environments in the United States and Japan (Torrens, 2012), and standard movement routines used in mathematical and physics models (Torrens et al., 2012). Participants in both experiments were asked to complete survey questionnaires.

    • Proof of concept study using GPS-based tracking data to build agent-based models of visitors’ off-trail behavior in nature-based tourism settings

      2022, Applied Geography
      Citation Excerpt :

      ABMs can compensate for this limitation as they allow for experimentation in situations where conducting actual experiments are undesirable or impossible or for assessing the impact of different scenarios (see Baggio & Baggio, 2020). Pedestrian models have been built using ABM tools, but the majority have examined urban-based phenomena such as wayfinding in cities, the dynamics of crowds in the built environment, and evacuation scenarios (Heppenstall et al., 2016; Johansson & Kretz, 2012; Torrens et al., 2012). Skov-Petersen (2008) predicted an increase in the use of ABM in recreation planning and visitor use management as technology improved.

    • Procedural crowd generation for semantically augmented virtual cities

      2021, Computers and Graphics (Pergamon)
      Citation Excerpt :

      In general, we can say our work presents a more powerful and flexible way of including semantics and defining complex relationships, allowing our environments to scale up, and easing the effort of creating diversity of behaviors for large crowds without much user intervention. In general, systems such as the ones presented by Katoshevski et al. [30], Rasouli and Timmermans [31], Torrens’ work [32,33], or the MATSim [34] project, randomly select locations and assign simple, fixed plans for the agents (e.g., Home-Work-Home) based on probability distributions. They cannot, thus, reach the flexibility and control presented by our system.

    • TripCube: A Trip-oriented vehicle trajectory data indexing structure

      2018, Computers, Environment and Urban Systems
    View all citing articles on Scopus
    View full text