Search

Purpose This study aims to propose a new simulation approach for a real-life large and complex crowd management which takes into account deep learning algorithm. Moreover, the proposed model also determines the crowd level and also sends an alarm to avoid the crowd from exceeding its limit. Also, the model estimates crowd density in the pictures through which the study evaluates the deep learning algorithm approach to address the problem of crowd congestion. Furthermore, the suggested model comprises of two main components. The first takes the images of the moving crowd and classifies them into five categories such as “heavily crowded, crowded, semi-crowded, light crowded and normal,” whereas the second one comprises of colour warnings (five). The colour of these lights depends upon the results of the process of classification. The paper is structured as follows. Section 2 describes the theoretical background; Section 3 suggests the proposed approach followed by convolutional neural network (CNN) algorithm in Section 4. Sections 5 and 6 explain the data set and parameters as well as modelling network. Experiment, results and simulation evaluation are explained in Sections 7 and 8. Finally, this paper ends with conclusion which is Section 9 of this paper. Design/methodology/approach This paper addresses the issue of large-scale crowd management by exploiting the techniques and algorithms of simulation and deep learning. It focuses on a real-life case study of Hajj pilgrimage in Saudi Arabia that exhibits intricate pattern of crowd management. Hajj pilgrimage includes performing Umrah along with hajj that involves several steps which is a sacred prayer of Muslims performed at different time span of the year. Muslims from all over the world visit the holy city of Mecca to perform Tawaf that is one of the stages included in the performance of Hajj or Umrah, it is an obligatory step in prayer. Accordingly, all pilgrims require visiting Mataf to perform Tawaf. It is essential to control the crowd performing Tawaf systematically in a constrained place to avoid any mishap. This study proposed a model for crowd management system by using image classification and a system of alarm to manage millions of people during Hajj. This proposed system highly depends on the adequate data set used to train CNN which is a deep learning technique and has recently drawn the attention of the research community as well as the industry in changing applications of image classification and the recognition of speed. The purpose is to train the model with mapped image data, making it available to be used in classifying the crowd into five categories like crowded, heavily crowded, semi-crowded, normal and light-crowded. The results produce adequate signals as they prove to be helpful in terms of monitoring the pilgrims which shows its usefulness. Findings After the first attempt of adding the first convolutional layer with 32 filters, the accuracy is not good and stands out at about 55%. Therefore, the algorithm is further improved by adding the second layer with 64 filters. This attempt is a success as it gives more improved results with an accuracy of 97%. After using the dropout fraction as a 0.5 to prevent overfitting, the test and training accuracy of 98% is achieved which is acceptable training and testing accuracy. Originality/value This study has proposed a model to solve the problem related to estimation of the level of congestion to avoid any accidents from happening because of it. This can be applied to the monitoring schemes that are used during Hajj, especially in crowd management during Tawaf. The model works as such that it activates an alarm when the default crowd limit exceeds. In this way, chances of the crowd reaching a dangerous level are reduced which minimizes the potential accidents that might take place. The model has a traffic light system, the appearance of red light means that the number of pilgrims in a particular area has exceeded its default limit and then it alerts to stop the migration of people to that particular area. The yellow light indicates that the number of pilgrims entering and leaving a particular area has equalized, then the pilgrims are suggested to slower their pace. Finally, the green light shows that the level of the crowd in a particular area is low and that the pilgrims can move freely in that area. The proposed model is simple and user friendly as it uses the most common traffic light system which makes it easier for the pilgrims to understand and follow accordingly.

This article explores whether crowd-sourced human creativity within a gamified collaborative design framework can address the complexity of predictive environment design. This framework is predicated on gamifying crowd objectives and presenting environment design problems as puzzles. A usability study reveals that the framework is considered usable for the task. Participants were asked to configure an environment puzzle to reduce an important crowd metric, the total egress time. The design task was constructed to be straightforward and uses a simplified environment as a probe for understanding the utility of gamification and the performance of collaboration. Single-player and multiplayer designs outperformed both optimization and expert-sourced designs of the same environment and multiplayer designs further outperformed the single-player designs. Single-player and multiplayer iterations followed linear and exponential decrease trends in total egress time, respectively. Our experiments provide strong evidence toward an interesting novel approach of crowdsourcing collaborative environment design.

At present, the traditional sports teaching model has been unable to meet the needs of modern diversified talents. Based on the actual needs of physical education, this article reforms traditional physical education methods, proposes a university physical education model based on virtual crowd simulation and digital media, and builds a corresponding system. Moreover, this paper uses feature identification to monitor classroom teaching and counts multiple parameters to assist the effective teaching activities. In model tracking, the DMS motion trajectory is obtained, and then the center points of all grid sequences are calculated to generate a center point matrix. Considering the difference in motion between adjacent frames of DMS, the difference is quantified by the significance value, so that the sequence of adjacent frames with small differences is organized into clusters. In addition, this paper builds a computer system model based on actual needs and evaluates model performance through actual teaching. The research results show that the proposed model has good performance.

Crowd simulations imitate the group dynamics of individuals in different environments. Applications in entertainment, security, and education require augmenting simulated crowds into videos of real people. In such cases, virtual agents should realistically interact with the environment and the people in the video. One component of this augmentation task is determining the navigable regions in the video. In this work, we utilize semantic segmentation and pedestrian detection to automatically locate and reconstruct the navigable regions of surveillance-like videos. We place the resulting flat mesh into our 3D crowd simulation environment to integrate virtual agents that navigate inside the video avoiding collision with real pedestrians and other virtual agents. We report the performance of our open-source system using real-life surveillance videos, based on the accuracy of the automatically determined navigable regions and camera configuration. We show that our system generates accurate navigable regions for realistic augmented crowd simulations.

In an emergency route planning technique, simulating the dynamic crowd has route capacity constraints and global target of evacuating all crowd evacuees. To stimulate the crowd, the new arena is developed to know the real-time situation to face the crowd evacuation on exit point. The crowd evacuation is done with the process of Hybrid Agent Reinforcement Learning (HARL) algorithm consisting of Improved Multi-Agent Reinforcement Learning (IMARL) and State-Action-Reward-State-Action (SARSA). In the proposed work, the appropriate route selection mechanism focused on finding optimum evacuation route(s) is done in the first phase. Dynamic crowd can also be evacuated to find its way with the support of the HARL process in the second phase. The proposed HARL method can also be implemented with multi-objective improved particle swarm optimization (IPSO) technique for crowd simulation. The experimented results demonstrate the effectiveness of stability in the HARL process, which provides an improved performance for crowd simulation

Agent based simulation (ABS) is a paradigm to modelling systems included of autonomous and interacting agents. ABS has been tremendous growth and used by researchers in the social sciences to study socio-environmental complex systems. To date, various platforms have been introduced for agent-based social simulation. They are rule based in any logic, python based in SPADE and etc. Although those platforms have been introduced, there is still an insufficient to develop a crowd simulation in 3D platform. Having a 3D platform is needed to enabling the crowd simulation for training purposes. However, the current tools and platform still lack features to develop and simulate autonomous agents in the 3D world. This paper introduced a BDI plug in at Unity3D for crowd simulation. BDI is an intelligent agent architecture and it is able to develop autonomous agents in crowd environment. In this paper, we present the BDI plug with a case study of Australia bush fire and discuss a method to support autonomous agents' development in 3D crowd simulation. The tool allows the modeller to develop autonomous agents in 3D world by taking the advantages of Unity3D.

In a crisis situation, negative emotions often spread among the crowd, and they have adverse impacts on human decisions, resulting in stampedes and crushes. Safety officers are often dispatched to scenes of emergencies because their positive emotions can calm the crowd down and avoid serious accidents. However, how to utilize the positive emotional contagion to maximize the “calm-down” effect remains a challenging problem in crowd evacuation. In this paper, we present an approach for optimizing positive emotional contagion in crowd evacuation. First, a computational model of positive emotional contagion is proposed to describe how safety officers calm a crowd down. To capture important influential factors for positive emotional contagion, such as the trust relationships among the individuals involved in a crisis situation and the variations of emotional contagion speed, we construct a trust-based emotional contagion network (Trust-ECN) and a heterogeneous emotional contagion speed computation model (HECS-CM). Based on these models, the emotional contagion process can be analyzed in a parametric way, and the infection probability for each individual in a given time window can be computed analytically with a continuous-time Markov chain (CTMC). Second, a maximization problem of emotional contagion is formulated. Since this optimization problem is NP-hard, an artificial bee colony optimized emotional contagion (ABCEC) algorithm is used to solve for the optimal positions of safety officers. We demonstrate the effectiveness of our method on both synthetic and real-world data at different scales. Finally, we implement a crowd simulation system to visualize the results of our theoretical analysis in a graphical manner. The proposed method can provide guidance for emergency response management.

Traditional methods of crowd evacuation simulation reduce the visual realism of crowd motion modeling due to hypothetical scenarios and rules. Data-driven methods are an effective way to enhance the visual realism of crowd simulation. However, the existing work mainly focuses on training models for specific scenarios and applies them to the same scenario, so it lacks consideration of scene adaptability. To address this problem, we present a residual network based scene-independent crowd simulation (ResNet-SICS) framework to simulate crowd motion. First, a data-driven crowd properties quantization (DCPQ) method is proposed. This method divides crowd properties into physical properties and social properties: the physical properties (such as the position and the velocity) are extracted from a large number of real videos, and the social properties (such as local cohesiveness and global collectiveness) are then quantified according to the physical properties of crowds. Second, a residual network for crowd behavior properties learning (ResNet-CBPL) motion prediction model is established. This model takes the crowd properties as parameters to construct the residual network, and uses real data to learn the movement rules of the crowd. The ResNet-CBPL motion prediction model can fit the movement behavior of the crowd more accurately than can other models. Finally, we implement a crowd simulation system based on ResNet-CBPL to visualize the results of theoretical analysis in a graphical way. The experimental results show that the proposed method can simulate the crowd motion more realistically, and the trained crowd simulation framework can be applied to various scenarios.

One limitation of current data-driven automatic crowd modeling methods is that the models generated have low interpretability, which limits the practical applications of the models. In this article, we propose a new data-driven crowd modeling approach that can generate universal behavior rules with better interpretability. Higher interpretability helps people better understand and analyze the rules. Furthermore, the proposed approach considers both static and dynamic features during modeling to generate a realistic crowd, based on the assumption that humans tend to consider different features with respect to their states. In the proposed method, the automatic behavior rule generation problem is formulated as a symbolic regression problem. Then, the problem is solved by multi-objective genetic programming. On one hand, to improve the interpretability of the behavior rules found, a new mechanism is proposed to guide the algorithm to find concise and dimensionally consistent solutions. On the other hand, decisions made by considering static and dynamic features respectively are combined to improve the generated crowd realism. To validate the effectiveness of the proposed method, three real-world datasets are utilized for training and testing. The simulation results demonstrate that the proposed method is able to find universal behavior rules that are competitive to previous work in accuracy while having better interpretability.

Evaluating and optimizing the design of built and yet-to-be-built environments, with respect to human occupancy and behavior is both greatly beneficial and challenging. Crowd simulation can provide the computational means to analyze a design through the movement of virtual occupants (agents). A range of analytic information (metrics) can be computed from the simulated movement of the agents that offer insights on the design. Crowd simulation and the related analysis can be part of interactive or offline design optimization pipelines. Unfortunately, large scale crowd simulations are prohibitively expensive, especially when used within iterative design and optimization loops, where hundreds of simulations often need to be computed at interactive rates. We propose a machine learning framework that aims to solve this problem by learning the relationship between a building design and the evaluation metrics extracted from expensive simulations. We train an offline regression neural network using a synthetic training set that we generate for this purpose. Once the network is trained it can evaluate new designs efficiently, and approximate the corresponding analytic information with high accuracy. The proposed framework can also be used to find an optimized layout. We demonstrate the effectiveness of the framework on a variety of real world case studies.

Modeling visual attention is an important aspect of simulating realistic virtual humans. This work proposes a parametric model and method for generating real-time saliency maps from the perspective of virtual agents which approximate those of vision-based saliency approaches. The model aggregates a saliency score from user-defined parameters for objects and characters in an agent’s view and uses that to output a 2D saliency map which can be modulated by an attention field to incorporate 3D information as well as a character’s state of attentiveness. The aggregate and parameterized structure of the method allows the user to model a range of diverse agents. The user may also expand the model with additional layers and parameters. The proposed method can be combined with normative and pathological models of the human visual field and gaze controllers, such as the recently proposed model of egocentric distractions for casual pedestrians that we use in our results.

In recent years, human trajectory prediction (HTP) has garnered attention in computer vision literature. Although this task has much in common with the longstanding task of crowd simulation, there is little from crowd simulation that has been borrowed, especially in terms of evaluation protocols. The key difference between the two tasks is that HTP is concerned with forecasting multiple steps at a time and capturing the multimodality of real human trajectories. A majority of HTP models are trained on the same few datasets, which feature small, transient interactions between real people and little to no interaction between people and the environment. Unsurprisingly, when tested on crowd egress scenarios, these models produce erroneous trajectories that accelerate too quickly and collide too frequently, but the metrics used in HTP literature cannot convey these particular issues. To address these challenges, we propose (1) the A2X dataset, which has simulated crowd egress and complex navigation scenarios that compensate for the lack of agent-to-environment interaction in existing real datasets, and (2) evaluation metrics that convey model performance with more reliability and nuance. A subset of these metrics are novel multiverse metrics, which are better-suited for multimodal models than existing metrics. The dataset is available at: https://mubbasir.github.io/HTP-benchmark/.

In outdoor crimes such as robbery and kidnapping, suspects generally secretly follow their victims in public places and then look for opportunities to commit crimes. Video anomaly detection (VAD) has achieved fruitful results through deep neural networks (DNN). However, as an abnormal behavior without obvious abnormal physical features, hidden following is highly similar to ordinary walking and accompanying behaviors, so it is difficult to effectively detect hidden dangerous followers using video anomaly detection methods or traditional trajectory analysis methods. We propose "hidden follower'' detection (HFD) task and a HFD model based on gaze pattern extraction. It extracts gaze pattern features of pedestrians from gaze-interval-series and introduces a time series classification model to classify pedestrians with or without hidden following purposes. Based on this model, we propose a hidden follower detection framework (HFDF) to detect hidden followers from normal pedestrians, which utilizes the trajectories and gaze patterns extracted from videos. To cope with the lack of test data, we construct a dataset of 1200 pedestrians from the crowd simulation model to simulate scenes including hidden followers, and we also collected a surveillance video dataset including the hidden following behaviors. The experiments conducted on these two datasets show that HFDF can consistently outperform the state-of-the-art method by a notable margin in the HFD task on the commonly-used F1 benchmark.

This paper presents an approach for the automatic abstraction of built environments needed for pedestrian dynamics from any building configuration. The approach assesses the usability of navigation mesh to perform realistically pedestrian simulation considering the physical structure and pedestrian abilities for it. Several steps are examined including the creation of a navigation mesh, space subdivision, border extraction, height map identification, stairs classification and parametrisation, as well as pedestrian simulation. A social-force model is utilised to simulate the interactions between pedestrians and an environment. To perform quickly different 2D/3D geometrical queries various spatial indexing techniques are used, allowing fast identification of navigable spaces and proximity checks related to avoidance of people and obstacles in built environments. For example, for a moderate size building having eight floors and a net area of 13,000 m2, it takes only 104 s to extract the required building information to run a simulation. This approach can be used for any building configuration extracting automatically needed features to run pedestrian simulations. In this way, architects, urban planners, fire safety engineers, transport modellers and many other users without the need to manually interact with a building model can perform immediately crowd simulations.

The paper suggests that one of the aspects of the pre-Covid world that is most missed in the ‘new normal’ is the social experience of crowds and that immersive media might compensate for this. The nature of crowds is explored with a view to identifying the parameters for successful crowd simulation. There is a brief survey of computer crowd simulation which identifies some of the challenges together with descriptions of some recent games which feature crowds as a significant element. The paper then describes a method using real-time searches of social media content derived from Twitter as an extended data set which can be used to produce variety in the individual agents making up generated crowds and so provide a closer simulation of the crowd experience.

Traditional methods of crowd evacuation simulation mainly focus on improving crowd evacuation efficiency while many assumptions reduce the visual realism of crowd evacuation. The data-driven method is an effective way to enhance the visual realism of crowd simulation whose main idea is to approach the trajectories of crowd movements and social attributes in real video as much as possible. However, the existing work lacks consideration of variable scenarios and dynamic movements which decreases the visual realism of crowd simulation. To address this problem, we present a reinforcement learning based data-driven crowd evacuation (RL-DCE) framework. In the proposed framework, a data-driven crowd evacuation (DCE) model is established first. The model extracts the dynamic characteristics (such as the position and the velocity) from videos to quantify the cohesiveness which is one of the most important social attributes of crowds. Then, we propose a cohesiveness based K-means (C-K-means) algorithm to group the crowd and merge the individual’s trajectories according to the groups. Second, a hierarchical path planning mechanism is proposed. In this mechanism, the top-layer employs the Q-learning algorithm (which is one of the reinforcement learning algorithms) to train the control policy using the obtained grouping trajectories. The Q-learning algorithm allows individuals to learn the main observed features of crowd motion (e.g., cohesiveness) and demonstrate robustness with respect to dynamic environment. The bottom-layer is mainly in charge of obtaining individual paths and using the reciprocal velocity obstacles (RVO) model to avoid collision. Finally, we implement a crowd simulation system based on the DCE model in dynamic environments. The experimental results show that the proposed method can simulate the crowd evacuation more realistically in the dynamic environment.

This paper presents the architecture, main components and performance results for a parallel and modular agent-based environment aimed at crowd simulation. The environment allows to simulate thousands or more agents on maps of square kilometers or more, features a modular design and incorporates non-volatile RAM (NVRAM) with a fail-safe mode that can be activated to allow to continue computations from a recently analyzed state in case of a failure. We show results for an evacuation scenario for an area of up to 6 km 2 in a district of Gdansk, Poland, performed on two clusters, one with hardware simulation of NVRAM. We have shown a very small overhead of using NVRAM compared to the RAM only solution and an overhead of 20% with the fail-safe mode on using NVRAM, shown up to 30 000 agents and up to 25 000 iterations of the simulation. We also show the benefit of using NVRAM for file synchronization with a slow growth of the execution time while increasing the map size. We then present how the frequency of visualization affects execution time and very good scaling of the proposed solution in a cluster environment for more than 650 processes and 60 000 agents.

The artificial bee colony (ABC) algorithm has the problem of slow convergence and may be trapped into local optimum. In this paper, a multi-species ABC (MABC) algorithm is proposed based on the multi-swarm model. The MABC algorithm uses dynamic segmentation of the swarm and a co-evolution strategy. The strategy of dynamic segmentation divides the colony into multiple sub-species, and the species communicate with each other using the co-evolution strategy. The combined global communication pattern and local communication pattern are applied among sub-species. In order to test the performance of the algorithm, experiments are conducted on the CEC’05 Test Functions. To test the performance of the algorithm in crowd simulation for evacuation, we simulate it through a crowd simulation system for evacuation, and the MABC algorithm improves the efficiency of crowd evacuation.

It is always difficult to evacuate crowds in public places like subway stations. The traditional crowd behavior simulation models often ignore two important issues in crowd evacuation: pedestrian tracking and individual differences. To solve the problem, this paper combines social force model (SFM) with deep learning into a novel pedestrian detection method. Firstly, several deep learning algorithms for pedestrian detection were compared, and the best ones for sparse and dense crowds were determined. Next, the pedestrian positions in a real video were acquired by the selected algorithms, and converted into actual coordinates in the scene. Then, the evacuation process was simulated with our method and the SFM based on these coordinates. The results show that our model output closer-to-reality results than the SFM. The research findings shed important new light on evacuation in crowded areas.

The combination of multi-agent technology and reinforcement learning methods has been recognized as an effective way which is used in path planning-based crowd simulation. However, the existing solution is still not satisfactory due to the problem in the mutual influence of agents. Therefore, an improved multi-agent reinforcement learning method (IMARL algorithm) is introduced. In this method, the intersection of the pedestrian trajectory extracted from the real video is first used as the state space for reinforcement learning. The crowd is grouped and the leader is selected. A bulletin board is added to the reinforcement learning algorithm of multi-agent to store the empirical knowledge of the learning process, and the navigation agent passes information between the leader and the bulletin board. The original social force model was improved, and the cohesive force of visual factors was added to the force formula. The IMARL algorithm is combined with the improved social force model for crowd evacuation simulation. Using a two-layer control mechanism, the leader in the upper layer uses the decision process based on the IMARL algorithm to select the path, and the individuals in the bottom group use the improved social force model to evacuate. The method of this paper not only solves the dimensionality disaster problem of reinforcement learning but also improves the convergence speed. The evacuation efficiency is effectively improved in crowd evacuation simulation experiments. In addition, it can also provide specific guidance scheme for crowd evacuation improvement and assistant decision support for the prevention and management of large-scale group trampling incidents.

Over the last two decades there has been a proliferation of methods for simulating crowds of humans. As the number of different methods and their complexity increases, it becomes increasingly unrealistic to expect researchers and users to keep up with all the possible options and trade-offs. We therefore see the need for tools that can facilitate both domain experts and non-expert users of crowd simulation in making high-level decisions about the best simulation methods to use in different scenarios. In this paper, we leverage trajectory data from human crowds and machine learning techniques to learn a manifold which captures representative local navigation scenarios that humans encounter in real life. We show the applicability of this manifold in crowd research, including analyzing trends in simulation accuracy, and creating automated systems to assist in choosing an appropriate simulation method for a given scenario.

In simulation studies of crowd behaviors, group behavioral modeling is the core issue of crowd simulation evacuation research. The core issue in group behavioral modeling is how to construct the group. This paper proposes examining the dissimilarities between pedestrians by tracing the trajectories of pedestrians from real videos and constructing groups under real scenes based on the dissimilarities between individuals according to the principle of common fate. The inter-group dissimilarity and intra-group dissimilarity are calculated by pedestrians’ trajectories and groups obtained. We present an improved social force model by the social comparison theory and intra-group attraction strategy combined with the data from video. In this paper, the scene information, the locations of the pedestrians, the groups of pedestrians, the intra-group dissimilarity and the inter-group dissimilarity are extracted from the real video. In addition, we considered intra-group attraction and the characteristic of the social comparison theory that lead pedestrians to gather together in emergency. Therefore, the method proposed in this paper has higher authenticity. To verify the effectiveness of the proposed method, we qualitatively and quantitatively designed comparative experiments, respectively. The experimental results show that the method can simulate the group behavior and reproduce the evacuation of people in emergency in real scenes. This model can serve as an important tool to predict evacuation statuses in different population densities and group distributions in emergency. It provides a basis for the establishment of a system to track pedestrians and predict their congestion.

Data-driven crowd evacuation learning methods are often used to enhance the realism of crowd simulation. However, the learning results of traditional methods cannot adapt to the dynamic changes of the simple scene, and thus have the disadvantage of poor generalization. To solve this problem, we propose a data-driven crowd evacuation framework based on hierarchical deep reinforcement learning. The framework consists of: a macro-control layer with path programming function and a micro control layer with collision avoidance function. In this paper, a path programming method combining data-driven and deep reinforcement learning is proposed in the macro-control layer. The method combines the pedestrian motion attributes in the video with the DDPG algorithm to learn the pedestrian track in the video from a macro perspective. In the micro-control layer, the track sequence learned in the macro-control layer is used as the motion target to learn the collision-free motion velocity of individuals using the multiple agent deep reinforcement learning method. When the scene changes, the micro-control layer adaptively adjusts the motion speed without the need for the macro-control layer to repeat the path programming learning. The experimental results demonstrate that the proposed hierarchical crowd evacuation framework can not only simulate the real crowd movement behavior and improve the simulation fidelity, but also flexibly adapt to the dynamic changes of the simple scene and enhance the generalization.

We use deep learning technique for predicting temporary target position of each controlling agent in crowd simulation. Our deep learning model is based on recurrent neural network, which can predict a heatmap that represents the surrounding situation of an agent from a series of previous heatmaps of the agent. A heatmap contains positions and speeds of nearby agents and the temporary target position which represents the heading direction of the agent. Each agent is controlled based on the temporary target position that is chosen from an estimated heatmap using a force-based model.

Risk-based security is a concept introduced in order to provide security checks without inconveniencing travelers that are being checked with unqualified scrutiny checks while maintaining the same level of security with current check point practices without compromising security standards. Furthermore, risk-based security, as a means of improving travelers’ experience at check points is expected to reduce queueing and waiting times while improving at the same travelers’ experience during checks. A number of projects have been funded by the European Commission to investigate the concept of risk-based security and develop the means and technology required to implement it. The author is the Coordinator of two of the flagship projects funded by EC on risk-based security: FLYSEC and TRESSPASS. This chapter discusses and analyses the concept of risk-based security, the inherent competing mechanism between risk assessment, screening time and level of security, and means to implement risk-based security based on anomaly detection using deep learning and artificial intelligence (AI) methods.

Existing crowd evacuation guidance systems require the manual design of models and input parameters, incurring a significant workload and a potential for errors. This paper proposed an end-to-end intelligent evacuation guidance method based on deep reinforcement learning, and designed an interactive simulation environment based on the social force model. The agent could automatically learn a scene model and path planning strategy with only scene images as input, and directly output dynamic signage information. Aiming to solve the &ldquo, dimension disaster&rdquo, phenomenon of the deep Q network (DQN) algorithm in crowd evacuation, this paper proposed a combined action-space DQN (CA-DQN) algorithm that grouped Q network output layer nodes according to action dimensions, which significantly reduced the network complexity and improved system practicality in complex scenes. In this paper, the evacuation guidance system is defined as a reinforcement learning agent and implemented by the CA-DQN method, which provides a novel approach for the evacuation guidance problem. The experiments demonstrate that the proposed method is superior to the static guidance method, and on par with the manually designed model method.

This paper presents a method for crowd motion segmentation and generating dominant motion patterns at the macroscopic crowd level, where a crowd is treated as an entity. In this approach, the dominant motion patterns, as a base for behaviour analysis of a mass of people, are the focus of interest. Dominant motion patterns are generated based on meta-trajectories. A meta-trajectory is defined as a set of tracklets and/or trajectories of entities in the crowd. The entities are particles initially organized as a uniform grid which is overlaid on a flow field. To estimate the flow field, a dense optical flow is used. Based on advection of the particles, tracklets/trajectories are obtained. They are grouped by a graph-based clustering algorithm and meta-trajectories are obtained. By overlapping meta-trajectories with the quantized orientation of the average optical flow field dominant motion patterns are obtained. The preliminary experimental results of the proposed method are given for a subset of UCF dataset, a subset of Crowd Saliency Detection dataset, our own FER dataset and computer crowd simulation videos of characteristic behaviour.

This paper presents a crowd simulation software that can assist building designers in visualizing and evaluating the building designs. It helps to simulate the movement of crowd in a building, visualize it and hence speed up decision-making process during building layout design stage. The software consists of 4 modules: the environment editor, the pathfinding module, the simulation module and the optimization module. The optimization engine assists the designers in improving their designs so as to ensure good trade-offs between various design parameters. In the example shown in this paper, the optimization feature is able to improve the evacuation process by 17.1% compared to the original design. It is envisaged that the crowd simulation software approach is critical in ensuring the safety of building occupants especially during emergency evacuation.

The topic of warehouse optimization is a critical and popular topic within the scientific literature [1, 4, 5, 7]. Particularly, order-picking optimization presents several profound difficulties: modeling, analysis, and data representation. Althoug these are necessary and obvious requirements for any research project, the visualization is an important aspect that has its challenges and complications. Menge is an open-source platform for crowd simulation. It provides an extensible common for research and development. In the present study, it is shown how the simulation platform known as Menge was adapted as a possible complement to the optimization process for the simulated visualization of its results. This chapter presents a solution using Menge to separate the modeling for the routing algorithm from visualization and simulation programming. The purpose of this work is to establish a starting point of industrial simulation in collaboration with algorithms for order-picking optimization in warehouses.

In large-scale events where many people gather, providing them with appropriate, efficient, and safe guidance about where to proceed is critical to ease congestion. We can evaluate guidance candidates using a pedestrian flow simulator to find appropriate guidance. However, evaluating many candidates by simulation requires high computational cost, which prohibits real-time guidance. We propose a method that finds appropriate guidance in real-time for observed situations based on deep reinforcement learning. Our proposed method learns a function that outputs appropriate guidance given the observed situation to minimize the average travel time of pedestrians. The difficulty here is that the real-world measurements of pedestrian travel time are limited due to privacy issues since it tracks individuals. Though our method uses only the observation obtained without locating specific individuals: the number of pedestrians who are moving on roads, it is guaranteed by Little’s law to be equivalent to minimizing the average travel time. Our experimental results for unknown pedestrian flow show that our proposed method outperforms rule-based controls, and its guidance is as effective as one selected from many candidates by repeated simulations with massive computational cost.

Risk-based and automatic security systems require to monitor passengers’ whereabouts in a terminal discretely to allow timely detection of suspicious behaviors and preventing malicious actions. In a series of two papers Thomopoulos et al. have introduced a methodology providing real-time risk assessment for airport passengers based on their trajectories. The proposed methodology implements a deep learning architecture. It is fully automated, reducing the workload of the video surveillance operators leading to less error-prone conclusions. Furthermore, the proposed methodology has been integrated with the OCULUS Command & Control (C2) System and the i-Crowd Simulator, a crowd simulation platform developed in the Integrated Systems Lab (ISL) of the Institute of Informatics and Telecommunications at NCSR “Demokritos.” In this paper we extend our previous work by introducing noise in both training and testing data used for tracking passengers and detecting anomalies in their tracks. Extensive testing of the anomaly detection system in the presence of noise demonstrates that the system is extremely resilient in noise. Furthermore, we consider the case of missing data in both training and testing data in order to model a realistic scenario of tracking with cameras with gaps in the passengers tracks from camera to camera due to missing data from transmission delays and/or data overflow. Extensive testing with the i- Crowd simulator demonstrates considerable robustness in the performance of the anomaly detection system in both noisy and missing data. The experimental results indicate that the proposed anomaly detection system is robust to both noisy and missing data and thus a very promising risk assessment scheme that can reliably be used for risk-based security under realistic operational conditions.

Crowd simulation is the act of simulating and controlling the dynamic movement of large groups of virtual characters. Crowd simulation is traditionally a complex and time-consuming process, requiring extensive manual effort to achieve. On the one hand, commercial and liberally licensed tools tend to have many aspects of simulation tightly integrated which can be prohibitively difficult to re-configure, on the other, paying extras can be far more costly. In this context, the re-use of existing simulated crowds has been identified as a valuable cost-saving approach to crowd simulations. Previous approaches have investigated the use of environment semantics, but they have not been integrated with a commonly used simulation platform, rendering their usefulness limited. We present a novel approach to crowd simulation using an emergent system for re-targeting autonomous crowds and report on the findings of a problem discovery study, analyzing and establishing key aspects of functionality, usability, and user experience. Our results provide a breakdown of the crowd simulation process with corresponding time-on-task metrics to provide a reference point for future scientific research into crowd simulation systems. Furthermore, we report on how users react to a system that involves the use of semantic data to facilitate the re-use of existing crowd simulations. We anticipate that other researchers will follow suit, to develop tools that are both innovative and usable in crowd simulation practices.

The safety of a dense crowd is one of the most important matters for an event’s organizers. Therefore, management of the crowd, and noticing any serious issues in advance becomes important. Developing a crowd simulation by using a social force model simulates the behavior of crowds in reality. The prediction of individual agents’ behavior in the simulation and how the agents interact with each other can improve the safety of dense crowds. Depending on the success of Recurrent Neural Network(RNN) handling of sequential data, we propose a model that is based on Long Short Term Memory (LSTM) to predict individual agents’ next locations. Our proposed approach will be tested two different densities of crowds, structured crowds, and unstructured crowds. In structured crowds, people generally move in one direction and head to the same destination, such as at the Islamic Hajj. In unstructured crowds people move in many different directions and head for different destinations, such as in public town squares.

The heterogeneity of the appearance and behavior of the crowd is an important component in a realistic simulation. Using Web3D technology to recreate a large-scale crowd of virtual avatars via the internet is an increasingly important area in crowd studies, especially for emergency crowd evacuation. One major issue in the early development of real-time simulation is the challenge of introducing diversity and authenticity on avatars’ appearance and behavior. Another major theoretical issue is high transmission delay on the network, which diminishes the simulation performed on the web browser when rendering a large number of visualization components. In this research, we propose a novel method to provide a lightweight solution that can simultaneously guarantee the realness and diversity of the simulated crowd. We use a parameterization technique based on shape space to distinguish the avatars’ appearance. Asynchronously transmitting the rendering elements helps to reduce bandwidth pressure. The multi-level clone instancing method can generate a massive amount of heterogeneous avatars in a short time. In the end, we validated our methods with an online experiment to demonstrate its ability to solve large-scale crowd simulation problems over the internet.

To simulate the low-level (‘microscopic’) behavior of human crowds, a local navigation algorithm computes how a single person (‘agent’) should move based on its surroundings. Many algorithms for this purpose have been proposed, each using different principles and implementation details that are difficult to compare. This paper presents a novel framework that describes local agent navigation generically as optimizing a cost function in a velocity space. We show that many state-of-the-art algorithms can be translated to this framework, by combining a particular cost function with a particular optimization method. As such, we can reproduce many types of local algorithms using a single general principle. Our implementation of this framework, named umans(Unified Microscopic Agent Navigation Simulator), is freely available online. This software enables easy experimentation with different algorithms and parameters. We expect that our work will help understand the true differences between navigation methods, enable honest comparisons between them, simplify the development of new local algorithms, make techniques available to other communities, and stimulate further research on crowd simulation.

Crowd behavior is the collective act and gathering of a group of individuals to achieve a shared purpose. Swarm intelligence-based optimization algorithms are usually used to solve complex problems for crowd behavior. Crowd simulations are often used for the analyses that require precision in different domains such as complex structural analysis, image recognition, creating nature-inspired non-player character movements in video games, and more. In this study, a generic crowd simulation framework that can be used to simulate already-available crowd simulation algorithms and design new ones was developed. The test environment layout was generated with the use of a generate-and-test algorithm combined with the crowd simulation algorithms to make sure that the generated content is meeting the requirements of a crowd simulation environment. Within the framework, three different crowd simulation algorithms —firefly algorithm, particle swarm optimization, and artificial bee colony— are generated and also implemented as puzzle-like video games. The results show that all fireflies achieved to gather at the global minimum of the generated layout faster and in a more precise way than the artificial bee colony algorithm and particle swarm optimization algorithm. The developed framework enables a generic and parametric testbed to design and compare different algorithms and to generate video games.

Dense crowds in public spaces have often caused serious security issues at large events. In this paper, we study the 2010 Love Parade disaster, for which a large amount of data (e.g. research papers, professional reports and video footage) exist. We reproduce the Love Parade disaster in a three-dimensional computer simulation calibrated with data from the actual event and using the social force model for pedestrian behaviour. Moreover, we simulate several crowd management strategies and investigate their ability to prevent the disaster. We evaluate these strategies in virtual reality (VR) by measuring the response and arousal of participants while experiencing the simulated event from a festival attendee’s perspective. Overall, we find that opening an additional exit and removing the police cordons could have significantly reduced the number of casualties. We also find that this strategy affects the physiological responses of the participants in VR., Journal of the Royal Society. Interface, 17 (167), ISSN:1742-5689, ISSN:1742-5662

We introduce a new inverse modeling method to interactively design crowd animations. Few works focus on providing succinct high-level and large-scale crowd motion modeling. Our methodology is to read in real or virtual agent trajectory data and automatically infer a set of parameterized crowd motion models. Then, components of the motion models can be mixed, matched, and altered enabling rapidly producing new crowd motions. Our results show novel animations using real-world data, using synthetic data, and imitating real-world scenarios. Moreover, by combining our method with our interactive crowd trajectory sketching tool, we can create complex spatio-temporal crowd animations in about a minute.

Crowd simulation is a central topic in several fields including graphics. To achieve high-fidelity simulations, data has been increasingly relied upon for analysis and simulation guidance. However, the information in real-world data is often noisy, mixed and unstructured, making it difficult for effective analysis, therefore has not been fully utilized. With the fast-growing volume of crowd data, such a bottleneck needs to be addressed. In this paper, we propose a new framework which comprehensively tackles this problem. It centers at an unsupervised method for analysis. The method takes as input raw and noisy data with highly mixed multi-dimensional (space, time and dynamics) information, and automatically structure it by learning the correlations among these dimensions. The dimensions together with their correlations fully describe the scene semantics which consists of recurring activity patterns in a scene, manifested as space flows with temporal and dynamics profiles. The effectiveness and robustness of the analysis have been tested on datasets with great variations in volume, duration, environment and crowd dynamics. Based on the analysis, new methods for data visualization, simulation evaluation and simulation guidance are also proposed. Together, our framework establishes a highly automated pipeline from raw data to crowd analysis, comparison and simulation guidance. Extensive experiments and evaluations have been conducted to show the flexibility, versatility and intuitiveness of our framework., accepted in SIGGRAPH 2020

Simulating a realistic navigation of virtual pedestrians through virtual environments is a recurring subject of investigations. The various mathematical approaches used to compute the pedestrians’ paths result, i.a., in different computation-times and varying path characteristics. Customizable parameters, e.g., maximal walking speed or minimal interpersonal distance, add another level of complexity. Thus, choosing the best-fitting approach for a given environment and use-case is non-trivial, especially for novice users.To facilitate the informed choice of a specific algorithm with a certain parameter set, crowd simulation frameworks such as Menge [3] provide an extendable collection of approaches with a unified interface for usage. However, they often miss an elaborated visualization with high informative value accompanied by visual analysis methods to explore the complete simulation data in more detail – which is yet required for an informed choice. Benchmarking suites such as SteerBench [14] are a helpful approach as they objectively analyze crowd simulations, however they are too tailored to specific behavior details. To this end, we propose a preliminary design of an advanced graphical user interface providing a 2D and 3D visualization of the crowd simulation data as well as features for time navigation and an overall data exploration.

To effectively carry out disaster prevention and mitigation, it is essential to study the special disaster risk system. When unexpected events occur in disaster cases with large-scale crowd, crowd evacuation is particularly important. Research on safety evacuation of large-scale crowd has been lasted for a long time. However, current simulation models focus on a closed simulation environment. That means the simulations cannot receive information and data from surroundings. Nevertheless, in real life, with the development of IoT and CPS, the crowd simulations need to dynamically adjust and correct their simulation steps, according to inputting data from sensors or analyzed information. Therefore, in this study, a novel CPS-aware crowd simulation framework is proposed. The framework can help simulation system to conduct and correct “better” results. The framework mainly consists of two components: (1) the simulation mode and (2) the feedback mode. The experimental results show that we can get better evacuation paths with our crowd simulation framework.

This work presents a novel First-person View based Trajectory predicting model (FvTraj) to estimate the future trajectories of pedestrians in a scene given their observed trajectories and the corresponding first-person view images. First, we render first-person view images using our in-house built First-person View Simulator (FvSim), given the ground-level 2D trajectories. Then, based on multi-head attention mechanisms, we design a social-aware attention module to model social interactions between pedestrians, and a view-aware attention module to capture the relations between historical motion states and visual features from the first-person view images. Our results show the dynamic scene contexts with ego-motions captured by first-person view images via FvSim are valuable and effective for trajectory prediction. Using this simulated first-person view images, our well structured FvTraj model achieves state-of-the-art performance.

Occupancy analyses represent a crucial topic for building performance. At present, this is even true because of the pandemic emergency due to SARS-CoV-2 and the need to support the functional analysis of building spaces in relation to social distancing rules. Moreover, the need to assess the suitability of spaces in high occupancy buildings as the educational ones, for which occupancy evaluations result pivotal to ensure the safety of the end-users in their daily activities, is a priority. The proposed paper investigates the steps that are needed to secure a safe re-opening of an educational building. A case study has been selected as a test site to analyse the re-opening steps as required by Italian protocols and regulations. This analysis supported the school director of a 2-to-10 year old school and its team in the decision-making process that led to the safe school re-opening. Available plants and elevations of the building were collected and a fast digital survey was carried out using the mobile laser scanner technology (iMMS - Indoor Mobile Mapping System) in order to acquire three-dimensional geometries and digital photographic documentation of the spaces. A crowd simulation software (i.e. Oasys MassMotion) was implemented to analyse end-users flows; the social distance parameter was set in its proximity modelling tools in order to check the compliance of spaces and circulation paths to the social distancing protocols. Contextually to the analysis of users flows, a plan of hourly air changes to maintain a high quality of the environments has been defined.

Population estimation typically involves counting a group of individuals at a specific place and time. Estimating people on the move, i.e. a mobile population, is an intellectual challenge both in ...

This paper addresses the problem of efficiently simulating a believable virtual crowd. Our method is the first one that uses the Neural Network (NN) model to fit behaviors from real crowd data to a crowd simulation. Unlike several rule-based approaches that often result in ‘walking robots’, our model can learn motion rules derived from real data and later simulate human walking motions. Additionally, unlike the existing data-driven crowd simulation methods that have to perform search operations on the bound dataset simultaneously during the simulation, our model directly uses the NN model to generate the proper motion for each crowd member. The proposed method is being tested on various scenarios and compared with state-of-the-art state-action-based methods that are commonly employed in data-driven crowd simulation systems. The results demonstrate a significant increase in speed, as well as better simulation quality.

In this article, we propose a role-dependent (RD) data-driven modeling approach to simulate pedestrians’ motion in high-density scenes. It is commonly observed that pedestrians behave quite differently when walking in a dense crowd. Some people explore routes toward their destinations. Meanwhile, some people deliberately follow others, leading to lane formation. Based on these observations, two roles are included in the proposed model: leader and follower. The motion behaviors of leader and follower are modeled separately. Leaders’ behaviors are learned from real crowd motion data using state-action pairs, while followers’ behaviors are calculated based on specific targets that are obtained dynamically during the simulation. The proposed RD model is trained and applied to different real-world datasets to evaluate its generality and effectiveness. The simulation results demonstrate that the RD model is capable of simulating crowd behaviors in crowded scenes realistically and reproducing collective crowd behaviors such as lane formation.

In this paper we describe a prototype implementation of an augmented reality (AR) system for accessing and interacting with crowd simulation software. We identify a target audience and tasks (access to the software in a science museum) motivate the choice of AR system (an interactive table complemented with handheld AR via smartphones) and describe its implementation. Our system has been realized in a prototypical implementation verifying its feasibility and potential. Detailed user testing will be part of our future work.

The occurrence of natural or man-made emergencies can be quite complex and demand flawless preparedness, through tested strategies, in order to ensure the safety of the individuals. For large-scale infrastructures, whether commercial or residential, having a reliable evacuation strategy is crucial. Formulation and evaluation of these evacuation strategies is however a daunting challenge. In this paper, we propose a Crowd Simulation and Analysis framework for the formulation and evaluation of effective evacuation strategies in large buildings, using real-scale building structures and agent based approach. We further propose an algorithm to devise an evacuation strategy. We first demonstrate the functionality of our algorithm using a simplistic example and then apply the algorithm in a campus evacuation case study using three scenarios. The main goal of this research is to assist regulatory authorities in developing effective disaster management plans through the use of M&S methods and tools.

Crowd simulations span a wide spectrum of application domains, most notably video games, evacuation scenarios, and the movie industry. However, it is not obligatory that all virtual populace applications have the primary objective of realistic simulation. In most instances, it is necessary and sufficient that viewers perceive the crowd as plausible. Even for a crowd consisting of agents navigating on linear trajectories without any maneuvers, visual motion illusion elicited by these trajectories might appear to be a natural consequence, causing them to be perceived as wriggling rather than straight. In this respect, we evaluate in this study whether simulated 3D human agents walking with constant, collision-free velocities, induce such a maneuvering motion illusion, aiming toward an efficient real-time crowd simulation. For this purpose, we recorded videos of virtual human crowds with different parameter combinations, such as the agent walking speed, crowd density, camera tilt angle, and camera distance. These videos were watched by human subjects who were instructed to mark the virtual agents who they thought had changed their gait directions. The analyzed results revealed that participants claimed the presence of maneuvering virtual agents in the videos, even though there were none in any of them. Spatial grouping of the markings highlighted that the participants mainly focused on the central area of the simulation environment, and spatiotemporal analysis of the click data also showed stronger evidence to such an illusion (see accompanying video). Furthermore, we found that all of the referred parameters have statistically significant main effects on the number of marked agents per watched video.