Crowd Motion Research Articles

Non-Intrusive Privacy-Preserving Approach for Presence Monitoring Based on WiFi Probe Requests

Monitoring the presence and movements of individuals or crowds in a given area can provide valuable insight into actual behavior patterns and hidden trends. Therefore, it is crucial in areas such as public safety, transportation, urban planning, disaster and crisis management, and mass events organization, both for the adoption of appropriate policies and measures and for the development of advanced services and applications. In this paper, we propose a non-intrusive privacy-preserving detection of people's presence and movement patterns by tracking their carried WiFi-enabled personal devices, using the network management messages transmitted by these devices for their association with the available networks. However, due to privacy regulations, various randomization schemes have been implemented in network management messages to prevent easy discrimination between devices based on their addresses, sequence numbers of messages, data fields, and the amount of data contained in the messages. To this end, we proposed a novel de-randomization method that detects individual devices by grouping similar network management messages and corresponding radio channel characteristics using a novel clustering and matching procedure. The proposed method was first calibrated using a labeled publicly available dataset, which was validated by measurements in a controlled rural and a semi-controlled indoor environment, and finally tested in terms of scalability and accuracy in an uncontrolled crowded urban environment. The results show that the proposed de-randomization method is able to correctly detect more than 96% of the devices from the rural and indoor datasets when validated separately for each device. When the devices are grouped, the accuracy of the method decreases but is still above 70% for rural environments and 80% for indoor environments. The final verification of the non-intrusive, low-cost solution for analyzing the presence and movement patterns of people, which also provides information on clustered data that can be used to analyze the movements of individuals, in an urban environment confirmed the accuracy, scalability and robustness of the method. However, it also revealed some drawbacks in terms of exponential computational complexity and determination and fine-tuning of method parameters, which require further optimization and automation.

Continual Pedestrian Trajectory Learning With Social Generative Replay

Learning to predict the trajectories of pedestrians is essential for improving safety and efficiency of mobile robots. The prediction is challenging since the robot needs to operate in multiple environments in which the motion patterns of pedestrians are different between environments. Existing pedestrian trajectory prediction models heavily rely on the availability of representative data samples during training. In the presence of additional training data from a new environment, these models must be retrained on all datasets to avoid <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">catastrophic forgetting</i> of the knowledge obtained from the already supported environments. In this paper, we address this catastrophic forgetting problem in the context of learning to predict the trajectories of pedestrians. We propose a pseudo-rehearsal approach based on a novel Generative Replay (GR) model, referred to as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Social-GR</i> . The proposed method is consistent with crowd motion patterns and is free of any explicit reference to past experiences. To demonstrate the problem of catastrophic forgetting and evaluate our solution, we develop the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Continual Trajectory Prediction Benchmark</i> , which consists of four tasks, each representing a real-world pedestrian trajectory dataset from a different environment. By conducting several experiments, we show that our proposed Social-GR approach significantly outperforms other continual learning methods that depend on explicit experience replay, including the state-of-the-art conditional-GR model. We further illustrate the robustness of our proposed approach to mitigating catastrophic forgetting by switching the order of environments and employing a more complex prediction model.

ViewBirdiformer: Learning to Recover Ground-Plane Crowd Trajectories and Ego-Motion From a Single Ego-Centric View

We introduce a novel learning-based method for view birdification [1], the task of recovering ground-plane trajectories of pedestrians of a crowd and their observer in the same crowd just from the observed ego-centric video. View birdification becomes essential for mobile robot navigation and localization in dense crowds where the static background is hard to see and reliably track. It is challenging mainly for two reasons; i) absolute trajectories of pedestrians are entangled with the movement of the observer which needs to be decoupled from their observed relative movements in the ego-centric video, and ii) a crowd motion model describing the pedestrian movement interactions is specific to the scene yet unknown a priori. For this, we introduce a Transformer-based network referred to as ViewBirdiformer which implicitly models the crowd motion through self-attention and decomposes relative 2D movement observations onto the ground-plane trajectories of the crowd and the camera through cross-attention between views. Most important, ViewBirdiformer achieves view birdification in a single forward pass which opens the door to accurate real-time, always-on situational awareness. Extensive experimental results demonstrate that ViewBirdiformer achieves accuracy similar to or better than state-of-the-art with three orders of magnitude reduction in execution time.

Spatial and temporal analysis of the bottleneck flow under different walking states with a moving obstacle

The regulation of a moving obstacle on crowd movement offers the possibility to enhance evacuation efficiency in emergency situations. In this paper, a series of controlled experiments are conducted to study the effect of the moving obstacle on crowd dynamics for pedestrians in three different competitive levels, which respectively correspond to three different walking states. The enhancement effects of the moving obstacle on evacuation efficiency for the crowd in the dual-task and high-motivated walking states are confirmed, and the positions of the moving obstacle are crucial. It is found that the moving obstacle diminishes the order of the trajectories for the crowd in the dual-task and normal walking states, while it boosts near the exit for the crowd in the high-motivated walking state. And the moving obstacle makes the linear backward propagations of stop-and-go wave disappear for the crowd in the dual-task and high-motivated walking states, but the frequency of stop behavior increases for the crowd in the dual-task and normal walking states. The profiles of evacuation time show that the moving obstacle impedes the pedestrian flow from the front of the exit and increases evacuation efficiency for the pedestrians near the walls of the exit. The analysis of time headway suggests that the moving obstacle with a gap of 1.0 m or 1.2 m to the exit can reduce the number of the pedestrians waiting near the exit for the crowd in the dual-task and high-motivated walking states. Besides, the gap of 0.8 m between the moving obstacle and the exit makes the conflicts at the exit is increased, but the gap of 1.0 m or 1.2 m makes the number of conflicts at the exit be reduced. This study helps the evacuation management of dense crowds and improves the design of facilities to facilitate pedestrian traffic.

Review of Virtual Simulation of Crowd Motion for Urban Emergency Management

With increasing urbanization in China, there are more and more urban crowds, bringing new challenges to public safety management. The analysis and modeling of crowd behavior are important aspects of emergency management in smart cities. This paper reviews and summarizes existing research into crowd behavior simulation and analyzes existing simulation data collection methods. Crowd simulation technology is discussed from three aspects: crowd modeling, motion navigation, and emotion-driven crowd animation. The indicators of the simulation models are analyzed from the perspective of model evaluation. Finally, combined with the status quo, the existing research limitations are analyzed, and the direction of further research into crowd simulation behavior is discussed.

Influences of obstacle factors on the transmission trends of respiratory infectious diseases in indoor public places

Public facilities are important transmission places for respiratory infectious diseases (e.g., COVID-19), due to the frequent crowd interactions inside. Usually, changes of obstacle factors can affect the movements of human crowds and result in different epidemic transmissions among individuals. However, most related studies only focus on the specific scenarios, but the common rules are usually ignored for the impacts of obstacles' spatial elements on epidemic transmission. To tackle these problems, this study aims to evaluate the impacts of three spatial factors of obstacles (i.e., size, quantity, and placement) on infection spreading trends in two-dimension, which can provide scientific and concise spatial design guidelines for indoor public places. Firstly, we used the obstacle area proportion as the indicator of the size factor, gave the mathematical expression of the quantity factor, and proposed the walkable-space distribution indicator to represent the placement factor by introducing the Space Syntax. Secondly, two spreading epidemic indicators (i.e., daily new cases and people's average exposure risk) were estimated based on the fundamental model named exposure risk with the virion-laden particles, which accurately forecasted the disease spreading between individuals. Thirdly, 120 indoor scenarios were built and simulated, based on which the value of independent and dependent variables can be measured. Besides, structural equation modeling was employed to examine the effects of obstacle factors on epidemic transmissions. Finally, three obstacle-related guidelines were provided for policymakers to mitigate the disease spreading: minimizing the size of obstacles, dividing the obstacle into more sub-ones, and placing obstacles evenly distributed in space.

Blurring Boundaries: Weakening 3rd-order Motion Reduces Locomotor Responses When Following A Crowd

Collective motion in human crowds can emerge from pedestrians locally following their neighbors. Previously we showed that the visual-motor system relies on 3rd-order motion (tracking boundary features), not 1st-order motion (motion energy) (Lu & Sperling, 1995), to follow a crowd. We used the reverse-phi illusion to put 1st- and 3rd-order motion in conflict. Here we show that weakening 3rd-order motion by blurring object boundaries reduces both heading and speed responses, when in conflict with 1st-order motion. Participants (n=12) wore a wireless head-mounted display (101ºH x 105ºV, 90 Hz) and were asked to “walk with” a crowd of virtual objects as if they are real people. The ‘crowd’ consisted of 9 rectangles (1.6mV x 0.5mH) mapped with a Julesz texture. The ‘crowd’ initially moved forward at 1.2 m/s, then was briefly perturbed (heading perturbation: ±20˚, translation left or right; speed perturbation: ±0.2 m/s, optical expansion or contraction). During the perturbation, object texture underwent constant 4-frame reverse-phi motion so 1st-order texture motion was in the same or opposite direction as 3rd-order boundary motion (Same/Opposite Condition). Object boundaries were blurred at three different levels (Zero, Small, Large). The participant’s walking direction and speed were recorded. In the Opposite condition, locomotor responses progressively decreased as object boundaries were more blurred (heading: 15.3˚, 13.2˚, 12.1˚; speed: 0.072, 0.041, 0.039 m/s, respectively), consistent with reliance on 3rd-order motion. In the Same condition, on the other hand, responses increased with more blur (heading: 16.4˚, 17.1˚, 18.3˚; speed: 0.105, 0.119, 0.121 m/s, respectively), likely because texture motion was faster than boundary motion. Linear mixed effects models found significant main effects and interactions for all comparisons. The results confirm that following a crowd is dominated by 3rd-order motion of bounded ‘figures’, although 1st-order motion can also influence responses when 3rd-order motion is weak.

Does attention influence who you follow in a crowd? Tracking neighbors vs. following your friends

Does attention increase the influence of neighbors in a crowd? Rio, Dachner, & Warren’s (2018) model assumes ensemble perception of crowd motion, and yet it seems that we can follow our friends and ignore others. We adapted the Multiple Object Tracking paradigm to test whether selective attention to a subset of the crowd spontaneously influences following behavior. In Experiment 1, the participant walked with a virtual crowd 9 neighbors, viewed in an HMD. At the beginning of each trial, a target set of 2-6 neighbors, or all 9 neighbors (control), blinked on and off; participants were instructed to attend to and track those neighbors as they walked with the crowd. The heading (walking direction) of targets or distractors was perturbed (±20°) midway through the trial. As a tracking check, a randomized memory probe appeared at the end of the trial. Participant heading, speed, facing direction (yaw angle) and probe responses were recorded. A linear mixed effects regression revealed that participant heading was influenced by the number of turning neighbors (χ2(1) = 17.19, p < 0.001), but not by attention to targets. Rather, participant heading was statistically equivalent to the average heading direction of the crowd (BF01 = 11.90), consistent with ensemble perception. Probe responses were highly accurate, indicating that participants successfully tracked targets. In Experiment 2, we tested whether participants can follow a group of “friends” when explicitly instructed to do so. Participants were first trained to recognize a set of avatar friends, which replaced blinking targets; otherwise the design was the same. In this case, participants fully segmented friends from a crowd of strangers and exclusively followed them (χ2(1) = 59.27, p<0.001), breaking ensemble perception. Subsequent experiments are investigating whether this “friend” effect is due to task-relevant attention (explicit instructions to follow targets) or grouping by familiarity.

A Fusion Crowd Simulation Method: Integrating Data with Dynamics, Personality with Common

AbstractThis paper proposes a novel crowd simulation method which integrates not only modelling ideas but also advantages from both data‐driven methods and crowd dynamics methods. To seamlessly integrate these two different modelling ideas, first, a fusion crowd motion model is developed. In this model the motion of crowd are driven dynamically by different forces. Part of the forces are modeled under a universal interaction mechanism, which describe the common parts of crowd dynamics. Others are modeled by examples from real data, which describe the personality parts of the agent motion. Second, a construction method for example dataset is proposed to support the fusion model. In the dataset, crowd trajectories captured in the real world are decomposed and re‐described under the structure of the fusion model. Thus, personality parts hidden in the real data could be locked and extracted, making the data understandable and migratable for our fusion model. A comprehensive crowd motion generation workflow using the fusion model and example dataset is also proposed. Quantitative and qualitative experiments and user studies are conducted. Results show that the proposed fusion crowd simulation method can generate crowd motion with the great motion fidelity, which not only match the macro characteristics of real data, but also has lots of micro personality showing the diversity of crowd motion.

Sketching Vocabulary for Crowd Motion

AbstractThis paper proposes and evaluates a sketching language to author crowd motion. It focuses on the path, speed, thickness, and density parameters of crowd motion. A sketch‐based vocabulary is proposed for each parameter and evaluated in a user study against complex crowd scenes. A sketch recognition pipeline converts the sketches into a crowd simulation. The user study results show that 1) participants at various skill levels and can draw accurate crowd motion through sketching, 2) certain sketch styles lead to a more accurate representation of crowd parameters, and 3) sketching allows to produce complex crowd motions in a few seconds. The results show that some styles although accurate actually are less preferred over less accurate ones.

Fundamental diagrams for straight movement of pedestrians in water

Flood is one of the most destructive disasters in the world with high frequency, which normally results in considerable casualties and economic losses. Understanding the basic laws of crowd movement in water is helpful for establishing effective evacuation guidelines under floods. However, there is a lack of empirical data and research on crowd movement in floods. In this study, a series of experiments were carried out in an 8 m long corridor in a swimming pool with the depth of 0.60 m. It is found that the free speed of pedestrians in water is 51.61% lower than that on land, and the male is +25% faster than the female. The speed-density relation shows two phases under the observed density range [0.30,2.20]ped/m2. The speed firstly decreases with the increasing density and then stays at 0.81 m/s when the density exceeds 1.50 ped/m2, whereas the flow increases monotonically with the increase of density. The influence of water on speed and flow is emphasized by comparison with the fundamental diagrams on land. In water, the lateral swaying amplitude of crowds decreases with increasing density, but the swaying frequency remains around 1 Hz. The swaying amplitude (the mean amplitude is 0.10 m) is 100% larger than walking on land (the mean amplitude is 0.05 m) under the speed range [0.50,1.10]m/s, and the crowd step frequency in water was 33.07% less than that on land. Our findings can be useful for the improvement of flood evacuation modeling and establishing evacuation guidelines.

Guiding crowds when facing limited compliance: Simulating strategies.

At traffic hubs, it is important to avoid congestion of pedestrian streams to ensure safety and a good level of service. This presents a challenge, since distributing crowds on different routes is much more difficult than opening valves to, for example, regulate fluid flow. Humans may or may not comply with re-directions suggested to them typically with the help of signage, loudspeakers, apps, or by staff. This remains true, even if they perceive and understand the suggestions. Yet, simulation studies so far have neglected the influence of compliance. In view of this, we complement a state-of-the-art model of crowd motion and crowd behavior, so that we can vary the compliance rate. We consider an abstracted scenario that is inspired by a metro station in the city of Munich, where traffic regulators wish to make some passengers abandon the obviously shortest route so that the flow evens out. We investigate the effect of compliance for two very simple guiding strategies. In the first strategy, we alternate routes. In the second strategy, we recommend the path with the lowest crowd density. We observe that, in both cases, it suffices to reroute a small fraction of the crowd to reduce travel times. But we also find that taking densities into account is much more efficient when facing low compliance rates.

Toward multi-granularity spatiotemporal simulation modeling of crowd movement for dynamic assessment of tourist carrying capacity

ABSTRACT Dynamic process simulation and prediction of crowd movement are effective approaches to understanding the complex human behavior system in GIScience. At present, obtaining full-sample individual trajectory data still faces challenges because of privacy and cost constraints, thereby resulting in difficulty solving geographic modeling problems that require full-sample individual data. In this paper, a general model for crowd movement simulation is proposed by taking the dynamic evaluation of tourist carrying capacity as an example. Such method is a multi-granularity coupling model, which considers behavioral process and spatiotemporal heterogeneity of tourists. First, a secrete event-based logic model of tourist behavior is proposed. Second, a social force-based inference method of tourist path is designed. Finally, the simulation and evaluation model of remaining spatial carrying capacity of tourists based on a behavioral dynamic system is achieved. In addition, the correctness and applicability of the model are demonstrated through a case study. The proposed model will positively affect time- and space-sharing analysis and assessment of crowd flow within a specific area of activity.

Estimating indoor crowd density and movement behavior using WiFi sensing

The fact that almost every person owns a smartphone device that can be precisely located is both empowering and worrying. If methods for accurate tracking of devices (and their owners) via WiFi probing are developed in a responsible way, they could be applied in many different fields, from data security to urban planning. Numerous approaches to data collection and analysis have been covered, some of which use active sensing equipment, while others rely on passive probing, which takes advantage of nearly universal smartphone usage and WiFi network coverage. In this study, we introduce a system that uses WiFi probing technologies aimed at tracking user locations and understanding individual behavior. We built our own devices to passively capture WiFi request probe packets from smartphones, without the phones being connected to the network. The devices were tested at the headquarters of the research sector of the Elm Company. The results of the analyses carried out to estimate the crowd density in offices and the flows of the crowd from one place to another are promising and illustrate the importance of such solutions in indoor and closed spaces.

System Identification for the design of behavioral controllers in crowd evacuations

Behavioral modification using active instructions is a promising interventional method to optimize crowd evacuations. However, existing research efforts have been more focused on eliciting general principles of optimal behavior than providing explicit mechanisms to dynamically induce the desired behaviors, which could be claimed as a significant knowledge gap in crowd evacuation optimization. In particular, we propose using dynamic distance-keeping instructions to regulate pedestrian flows and improve safety and evacuation time. We investigate the viability of using Model Predictive Control (MPC) techniques to develop a behavioral controller that obtains the optimal distance-keeping instructions to modulate the pedestrian density at bottlenecks. System Identification is proposed as a general methodology to model crowd dynamics and build prediction models. Thus, for a testbed evacuation scenario and input–output data generated from designed microscopic simulations, we estimate a linear AutoRegressive eXogenous model (ARX), which is used as the prediction model in the MPC controller. A microscopic simulation framework is used to validate the proposal that embeds the designed MPC controller, tuned and refined in closed-loop using the ARX model as the Plant model. As a significant contribution, the proposed combination of MPC control and System Identification to model crowd dynamics appears ideally suited to develop realistic and practical control systems for controlling crowd motion. The flexibility of MPC control technology to impose constraints on control variables and include different disturbance models in the prediction model has confirmed its suitability in the design of behavioral controllers in crowd evacuations. We found that an adequate selection of output disturbance models in the predictor is critical in the type of responses given by the controller. Interestingly, it is expected that this proposal can be extended to different evacuation scenarios, control variables, control systems, and multiple-input multiple-output control structures.

Detecting Undeclared-Leader-Follower Structure in Pedestrian Evacuation Using Transfer Entropy

In evacuation models, individual interactions are typically assumed to follow the certain rules which are decided using prior knowledge. Our proposed improvement in evacuation modelling is inspired by the real-life experience that during crowd movements, individuals usually place their visual focus on the movements of their neighbours, even in the absence of social affiliations. That is, undeclared leader–follower (ULF) structures tend to emerge in crowd motion. In this study, to clarify the mechanism underlying individual interactions in crowd motion, a force-based model integrated with the ULF structure (LFM) was developed. An information-theoretic and model-free method, transfer entropy (TE), was applied to measure the ULF structure in evacuation crowds. Movement information (e.g., velocity and acceleration) was used as the time series for computing TE. The results showed that the LFM provides more realistic trajectories than does the classical social force model. In addition, the leader–follower structures were found to change over time, and an individual could act as a leader and a follower simultaneously during evacuation. ULF behaviour is strong at the early stage of evacuation and becomes weak when evacuees’ differences in movement states diminish. Moreover, the density map of leaders’ distributions presented a ‘V-like’ formation. Recognising the leader–follower relationship is central to understanding the mechanisms underlying the emergence of collective behaviour. The proposed approach is expected to aid in simulating realistic local interactions among evacuees and in providing a data-driven perspective for understanding crowd evacuation.

Analyzing the Impact of COVID-19 Control Policies on Campus Occupancy and Mobility via WiFi Sensing

Mobile sensing has played a key role in providing digital solutions to aid with COVID-19 containment policies, primarily to automate contact tracing and social distancing measures. As more and more countries reopen from lockdowns, there remains a pressing need to minimize crowd movements and interactions, particularly in enclosed spaces. Many COVID-19 technology solutions leverage positioning systems, generally using Bluetooth and GPS, and can theoretically be adapted to monitor safety compliance within dedicated environments. However, they may not be the ideal modalities for indoor positioning. This article conjectures that analyzing user occupancy and mobility via deployed WiFi infrastructure can help institutions monitor and maintain safety compliance according to the public health guidelines. Using smartphones as a proxy for user location, our analysis demonstrates how coarse-grained WiFi data can sufficiently reflect the indoor occupancy spectrum when different COVID-19 policies were enacted. Our work analyzes staff and students’ mobility data from three university campuses. Two of these campuses are in Singapore, and the third is in the Northeastern United States. Our results show that online learning, split-team, and other space management policies effectively lower occupancy. However, they do not change the mobility for individuals transitioning between spaces. We demonstrate how this data source can be a practical application for institutional crowd control and discuss the implications of our findings for policymaking.

A new approach to dominant motion pattern recognition at the macroscopic crowd level

Automatic analysis and the recognition and prediction of the behaviour of large-scale crowds in video-surveillance data is a research field of paramount importance for the security of modern societies. It serves to predict and help prevent disasters in public places where crowds of people gather. The paper proposes a novel method for generating meta-tracklets and recognition of dominant motion patterns as a basis for automatic crowd behaviour analysis at the macroscopic level, where a crowd is treated as an entity. The basic characteristic of macroscopic crowd scenes is that it is impossible to detect and track individuals in the scene. The idea of the method proposed in this paper is to recognize dominant crowd motion patterns, by avoiding time-consuming and error-sensitive crowd segmentation, crowd tracking and detection of regions of interest. Thus, the process of determining dominant motion patterns and recognizing crowd behaviour is accelerated. The method is inspired by a quantum mechanical approach. It combines a set of particles, which are considered as particles in quantum mechanics, tracklets of particles’ advection in a video clip, and the interaction of wave functions spread out from particle positions. A wave function is expressed in the form of an asymmetric potential function. Peaks of the wave field define the most probable particle flow, which defines a meta-tracklet. Dominant motion patterns are recognized by applying the functions of fuzzy predicates, which represent a combination of common-sense and human expert knowledge about crowd motions, to the meta-tracklets. The experimental results of the proposed method are presented for a subset of UCF dataset and AGORASET crowd simulation videos and have shown promising results in dominant motion pattern recognition.

Static floor field construction and fine discrete cellular automaton model: Algorithms, simulations and insights

While cellular automaton (CA) models have a broad range of applications in the field of pedestrian dynamics, there are still three significant issues (the efficient and accurate construction of the static floor field (SFF), the insufficient simulation accuracy, and the discretization effect) that have not been resolved or fully clarified yet. This work aims to tackle these challenges. First, a novel approximate algorithm that resolves the dilemma between accuracy and efficiency is proposed to construct the SFF, based on which the static navigation field is created by the proposed method. Then, a novel fine discrete CA model based on the desired direction is developed to capture the locomotion movement behaviour. The novel SFF construction algorithm and locomotion movement model are (theoretically and/or numerically) validated and compared to the state-of-the-art approaches in manifold scenarios. Later on, systematic simulation analyses are conducted to investigate the discretization effect at the micro and macro levels. The assessment metrics show that the performance of the proposed algorithm and model is superior over that of the state-of-the-art approaches. The proposed model can resolve the well-known diagonal movement artefact in the existing CA models and reproduce crowd movement and self-organized lane formation observed empirically. It is found that the non-isotropic SFF results in significantly biased and unrealistic evacuation movement. Moreover, the discretization degree is found to significantly affect the individual and crowd movement simulation outcomes, and the magnitude and direction of its effect depend on the used model and the setup of the scenarios. The proposed algorithm and model are promising tools in pedestrian modelling and simulation. The findings provide fresh insights into the development and application of CA models.

Modelling the dynamic relationship between spread of infection and observed crowd movement patterns at large scale events

Understanding how contact patterns arise from crowd movement is crucial for assessing the spread of infection at mass gathering events. Here we study contact patterns from Wi-Fi mobility data of large sports and entertainment events in the Johan Cruijff ArenA stadium in Amsterdam. We show that crowd movement behaviour at mass gathering events is not homogeneous in time, but naturally consists of alternating periods of movement and rest. As a result, contact duration distributions are heavy-tailed, an observation which is not explained by models assuming that pedestrian contacts are analogous to collisions in the kinetic gas model. We investigate the effect of heavy-tailed contact duration patterns on the spread of infection using various random walk models. We show how different types of intermittent movement behaviour interact with a time-dependent infection probability. Our results point to the existence of a crossover point where increased contact duration presents a higher level of transmission risk than increasing the number of contacts. In addition, we show that different types of intermittent movement behaviour give rise to different mass-action kinetics, but also show that neither one of two mass-action mechanisms uniquely describes events.