An extensible simulation environment and movement metrics for testing walking behavior in agent-based models
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)
- et al.
Modeling and prediction in a complex world
Futures
(2005) - et al.
Cellular automata microsimulation for modeling bi-directional pedestrian walkways
Transportation Research Part B
(2001) - et al.
Revealing the physics of movement: Comparing the similarity of movement characteristics of different types of moving objects
Computers, Environment and Urban Systems
(2009) - et al.
Urban layout, landscape features and pedestrian usage
Landscape and Urban Planning
(2007) - et al.
A microsimulation model for pedestrian flows
Mathematics and Computers in Simulation
(1985) - et al.
Macroscopic effects of microscopic forces between agents in crowd models
Physica A
(2007) - et al.
Statistical characteristics of evacuation without visibility in random walk model
Physica A: Statistical Mechanics and its Applications
(2004) - et al.
High quality navigation in computer games
Science of Computer Programming
(2007) - et al.
Crowd dynamics on a moving platform: Mathematical modelling and application to lively footbridges
Mathematical and Computer Modelling
(2007) - Allbeck, J., Kipper, K., Adams, C., Schuler, W., Zoubanova, E., Badler, N., et al. 2002. ACUMEN: amplifying control and...
Animal search strategies: A quantitative random-walk analysis
Ecology
Agent-based pedestrian modelling
The discrete dynamics of small-scale spatial events: Agent-based models of mobility in carnivals and street parades
International Journal of Geographical Information Science
Safety in numbers? Modelling crowds and designing control for the Notting Hill carnival
Urban Studies
Directing the herd: crowds and the science of evacuation
Science
The scaling laws of human travel
Nature
Pedestrian behaviour in pedestrian spaces: some findings of a naturalistic field study
Understanding the social context of the Schelling segregation model
Proceedings of the National Academy of Sciences
Collective behavior: An examination of some stereotypes
Social Problems
Agent street: An environment for exploring agent-based models in second life
Journal of Artificial Societies and Social Simulation
How we avoid collisions with stationary and moving obstacles
Psychological Review
Beauty is power: the use of space on the sidewalk
Sociometry
A note on two problems in connection with graphs
Numerische Mathematik
Spatial dynamics of the 1918 influenza pandemic in England, Wales and the United States
Journal of The Royal Society Interface
Modeling civil violence. An agent-based computational approach
Proceedings of the National Academy of Science
Coupled contagion dynamics of fear and disease: Mathematical and computational explorations
PLoS ONE
Crowd behaves as excitable media during Mexican wave
Nature
Strategies for mitigating an influenza pandemic
Nature
Pedestrian planning and design
Distance minimization in downtown pedestrian shopping
Environment and Planning A
Relations in public: Microstudies in the public order
Wayfinding behavior: Cognitive mapping and other spatial processes
Understanding individual human mobility patterns
Nature
Agent-based modeling of drinking behavior: A preliminary model and potential applications to theory and practice
American Journal of Public Health
Travelling waves and spatial hierarchies in measles epidemics
Nature
“PLEdestrians: A least-effort approach to crowd simulation
“So go downtown”: Simulating pedestrian movement in town centres
Environment and Planning B
A formal basis for the heuristic determination of minimum cost paths
IEEE Transactions on Systems, Science and Cybernetics
A fluid-dynamic model for the movement of pedestrians
Complex Systems
Social force model for pedestrian dynamics
Physical Review E
Self-organization phenomena in pedestrian crowds
Self-organizing pedestrian movement
Environment and Planning B
The statistics of crowd fluids
Nature
Disease in metapopulation models: implications for conservation
Ecology
The social logic of space
Gas-kinetic modeling and simulation of pedestrian flows
Transportation Research Record
Edestrian behavior at bottlenecks
Transportation Science
The flow of human crowds
Annual Review of Fluid Mechanics
Cited by (62)
Modeling of pedestrian turning behavior and prediction of pedestrian density distribution
2023, Physica A: Statistical Mechanics and its ApplicationsInverse augmentation: Transposing real people into pedestrian models
2023, Computers, Environment and Urban SystemsCitation 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 GeographyCitation 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.
Deep learning enabled Lagrangian particle trajectory simulation
2020, Journal of Aerosol ScienceTripCube: A Trip-oriented vehicle trajectory data indexing structure
2018, Computers, Environment and Urban Systems