Pedestrian Simulation Research Articles

Contact Risk Assessment in Dynamic Indoor Settings through Agent‐Based Modeling: A Spatially Explicit and Reproducible Approach

This study introduces an agent‐based model (ABM) pedestrian simulation tool to assess the risk of close contact (6 feet) in dynamic indoor environments, specifically in urban settings with diverse social activities and spatial structures. Our approach uses machine learning‐based sensitivity analysis (SA) to identify factors impacting the number of individual contacts, such as individual stay time and area. In addition, we conducted an in‐depth quantitative analysis to evaluate how specific factors, such as the strategic placement of obstacles, dwell time, and stay time near the entrances, mitigate the number of contacts. This analysis provides valuable insights for developing practical guidelines to curb contact risks in indoor environments. Lastly, we share the model, validation methods, and associated data as an open‐source Python library, complete with comprehensive documentation. This aims at fostering collaborative research and enables the application of our model across various scenarios, contributing to the development of spatially explicit models. Such efforts enhance the understanding of contact risks in urban indoor settings and promote joint research efforts, thus advancing the field through shared knowledge and tools.

Study on pedestrian evacuation simulation model considering group behavior

As a crowded place on the university campus, the cafeteria inevitably has many safety hazards, such as fire accidents caused by open flames and large appliances, stampede accidents caused by overcrowding during peak dining hours, etc. Therefore, studying pedestrian evacuation in university campus cafeterias is particularly necessary. Pedestrians on campus mostly travel in groups. Previous studies mainly used continuous models to discuss pedestrian group evacuation behavior. In this study, based on the cellular automaton pedestrian evacuation simulation model, the floor field calculation method was improved. A cellular automaton pedestrian evacuation simulation model considering group behavior was established and applied to the evacuation scenario of students in a university campus cafeteria. The study found that under the different group configurations, the pedestrian ratios, and the pedestrian densities, the pedestrian evacuation efficiency had significant differences. The results showed that when the different group configurations existed in the scenario, the higher the proportion of the pedestrians with three-person front-to-back group configurations, the higher the evacuation efficiency. When only one type of the group configuration existed in the scenario, at low pedestrian density, the evacuation efficiency of the individual pedestrian groups was higher compared to the other six group configurations. While at high pedestrian density, the evacuation efficiency of the three-person front-to-back group configurations was higher. These findings provided important references for pedestrian evacuation in university campus cafeterias and provided insights for the simulation research of group pedestrian evacuation models, contributing to enhancing campus safety management and ensuring the safety of teachers, students, and staff.

A bidirectional pedestrian macroscopic speed model construction based on pedestrian microscopic simulation experiments.

Studies of macroscopic speed modeling of bidirectional pedestrian cross-flows have relied heavily on scenario experiments, but the data itself may be deficient because large-scale scenario experiments are not easy to organize and subjects may not be walking under normal conditions. In order to explore the possibility of using microscopic pedestrian flow simulations for macroscopic speed modeling of pedestrian flows, a series of two-way pedestrian cross-flow simulation experiments were designed. Bidirectional pedestrian flows are defined as Peds1 and Peds2. The crossing angle and pedestrian flow rate are used as variables, and a bidirectional pedestrian flows simulation is designed as an orthogonal experiment. The crossing angles range from 15 to 165 degrees, and bidirectional pedestrian flow rate range from 1 ped/s to 8 ped/s. A series of simulations are built and performed on the GIS agent-based modeling architecture (GAMA) platform. By analyzing the flow data of bidirectional flows in the crossing area, it is found that when the Peds1 density falls below a threshold, Peds1 speed is determined by pedestrians themselves and mainly remains in a free flow state; otherwise, the Peds1 speed decreases with density. The clear effects such as Peds2 density on the Peds1 speed cannot be determined. A piecewise function combined with a linear function and an exponential function is constructed as the Peds1 speed model considering the influence of the crossing angle. The calibration results show that the piecewise function should be better than the non-piecewise function. Compared to the results of established studies, the results in this paper have some differences. Therefore, the simulation method cannot completely replace the scene experiments. However, this approach can provide suggestions for subsequent refinement of the experimental program, as well as a feasible direction for the construction of a speed relationship for bidirectional pedestrian flows.

A voxel-based 3D indoor model to support 3D pedestrian evacuation simulations

Three-dimensional (3D) indoor models enable realistic simulation of pedestrian evacuations in 3D space. However, existing indoor models do not fully identify navigable spaces located above/below physical components (e.g., chairs, stairs, desks), and the connections between these spaces and their navigable surfaces remain underexplored. These limitations hinder the accurate modelling of 3D pedestrian motions during evacuations. To address the shortcomings, we present novel automatic approaches to create a voxel-based 3D indoor model. First, based on different types of 3D motions, indoor spaces are automatically identified as freely navigable spaces, navigable spaces under conditions and non-navigable spaces. Then, navigable surfaces are identified to simulate moving above/below in 3D space. Vertical links are identified to connect the spaces and surfaces at different heights. A 3D pathfinding approach is subsequently developed to validate the model’s effectiveness. Finally, we evaluate the created model’s accuracy, robustness and efficiency across different voxel sizes. The primary findings are: 1) as voxel size increases, accuracy decreases, robustness changes and efficiency improves; 2) with very small voxel sizes, identification becomes challenging, while larger voxel sizes prevent some spaces and vertical links from being identified; and 3) indoor spaces with complex geometries require more time for identification. Our key contributions are: 1) an approach for automatically identifying indoor spaces for 3D motions; 2) two approaches for automatically identifying navigable surfaces and vertical links; and 3) a 3D pathfinding approach considering distinct 3D motion costs. This work advances 3D modelling to facilitate more realistic evacuation simulations.

Cellular automaton simulations of hybrid pedestrian movement in two-route situation

In this paper, we simulate the hybrid pedestrian movement considering both operational and tactical levels. The video data of our pedestrian route choice experiments are used for validations. Based on the CA model proposed by Nowak and Schadschneider, we build a new modeling framework and introduce six new parameters. Their values are determined by sensitivity analysis. To consider the influence of five statistical results in the simulations, including the velocities and densities of two routes, and the proportion of choosing the longer route, we propose a new metric named OPE (Overall Percentage Error). In order to reduce OPE, four models with different configurations are discussed and compared. We find that the use of shorter time step, modified SFF and the two-stage decision can help to improve the model performance. The simulation results in some other experimental run also shows the validity and universality of the model. In addition, we also compare the outputs of some typical pedestrian simulation software (e.g., AnyLogic), and we find our model has better results.

The Parameter Calibration of Social Force Model for Pedestrian Flow Simulation Based on YOLOv5.

With the increasing importance of subways in urban public transportation systems, pedestrian flow simulation for supporting station management and risk analysis becomes more necessary. There is a need to calibrate the simulation model parameters with real-world pedestrian flow data to achieve a simulation closer to the real situation. This study presents a calibration approach based on YOLOv5 for calibrating the simulation model parameters in the social force model inserted in Anylogic. This study compared the simulation results after model calibration with real data. The results show that (1) the parameters calibrated in this paper can reproduce the characteristics of pedestrian flow in the station; (2) the calibration model not only decreases global errors but also overcomes the common phenomenon of large differences between simulation and reality.

Performance Evaluation of Ekantakuna Intersection

Majority of the intersections in Kathmandu valley are unsignalized and hence traffic management becomes one of the emerging issues for traffic police and traffic designers. The number of vehicles in Kathmandu Valley is increasing at the rate of 12% annually (Department of Transport Management, 2018). The manual traffic management is very tedious and hectic nowadays. The consequence leads to traffic congestion and sometimes crashes too. Hence, the performance evaluation of such unsignalized intersection is essential and considering the real time situations, scenario analysis should be done. The research was carried out in Ekantakuna Intersection. The data's were collected in Ekantakuna Intersection through video-graphic surveying during weekdays: Monday to Friday both morning and evening during peak hours (9 to 11 AM in morning and 4 to 6 PM in evening).The model was developed in PTV VISSIM ( Micro-simulation software ) and thus the developed model was calibrated and validated using traffic volume. The relationship between field volume (actual volume) and VISSIM volume (Simulated Volume) was found with the help of R2 Value (Coefficient of Determination) which is found to be 0.9993 which indicates that variance of VISSIM output explains 99.9 percentage of variance in field data's. Pedestrian movement was not given priority and hence pedestrian simulation is not considered. Total three scenario analysis were performed in base case scenario. Among them the best scenario analysis was found to be scenario 1 which indicates the restriction of vehicles from service lanes to main lanes and allowing the movement of vehicles from Balkhu(service) to Satdobato(service) and Jawalakhel lanes, and Satdobato(service) to Balkhu(service ) and Bhaisepati lane.

Assessing and Improving Fire Safety Measures of Healthcare Facilities with Special Reference to Pedestrian Flow Evacuation Simulation and Evacuation Dynamics

Assessing and Improving Fire Safety Measures of Healthcare Facilities with Special Reference to Pedestrian Flow Evacuation Simulation and Evacuation Dynamics

Correlation Study of Commercial Street Morphology and Pedestrian Activity in Cold Region Summers under Thermal Comfort Guidance: A Case Study of Sanlitun, Beijing

Pedestrian vitality in commercial streets is influenced by various factors, among which the spatial form of the street and the resulting thermal environment have a significant impact. This study, from the perspective of thermal comfort, combines thermal comfort simulation with pedestrian simulation to establish an optimization model based on pedestrian vitality. The model aims to analyze and quantify the impact of street spatial form on thermal comfort and pedestrian vitality, providing a comprehensive evaluation of optimization schemes for commercial street spaces. Firstly, the study identifies the levels of spatial design parameters for commercial streets and generates optimized design scenarios for commercial street spaces. Using the simulation platforms Rhino 7 Grasshopper and MATLAB R2023a, a pedestrian simulation model guided by thermal comfort is constructed and validated against empirical data. Next, the influence of commercial street spatial design parameters on store visitations is assessed, identifying the most critical design parameters. Finally, design strategies for commercial streets are proposed based on vitality-oriented layouts. The results indicate that the spatial form of the street significantly affects store visitations, with the street width-to-height ratio being the most influential factor, followed by street orientation and interface form. NW-SE-oriented streets show a 47.2% higher Total Store Visitations (TSV) value compared to E-W-oriented streets, while E-W streets exhibit a Differential Store Visitation (DSV) value 4.47 times that of NW-SE streets. Streets with a W/H ratio of 0.25 have a 54.9% higher Total Store Visitations value than those with a W/H ratio of 0.9, and streets with a W/H ratio of 0.65 exhibit a Differential Store Visitations value 1.21 times that of streets with a W/H ratio of 0.25. Considering overall street vitality, the study recommends NW-SE- and NE-SW-oriented streets, with a width-to-height ratio between 0.25 and 0.4. The study also proposes strategies for the modification and expansion of streets in different orientations, providing the scientific basis and optimization recommendations for the planning and renovation of commercial streets in cold regions during summer.

The Birth of a New BIM Standard: From PED 2018 to 2023, New Parameters and Workflows "Going Live" for Everyone

Building Information Modelling (BIM) has become the de facto standard for the digital representation of buildings. However, from the pedestrian dynamics perspective, BIM Industry Foundation Classes (IFC) schema specification do not fully support data properties required for two-way data sharing with pedestrian modelling tools. An international team of academic and industry researchers, supported by buildingSMART International (bSI), is developing an Occupant Movement Analysis (OMA) standard. The project is focused on expanding the IFC schema specification to support workflows for pedestrian simulation tools and is close to completion. So far, multiple process maps and a list of data properties synchronised with several representative pedestrian modelling tools have been produced. This list of data properties was then converted into bSI's recommended flexible and machine interpretable Information Delivery Specification (IDS) format for specifying data exchange requirements and to add clarity. Currently, this is undergoing testing and review by the project team. Once completed, it will be submitted to bSI’s committees for review. Also, to support this work, a prototype open-source Add-in has been developed to demonstrate a two-way integrated data sharing between BIM authoring tools and pedestrian simulation tools. This standard will enhance data sharing between BIM authoring and pedestrian modelling tools by facilitating the capturing of the required data and addressing friction in multiple design iterations and reassessment.

Children’s ability to estimate approaching vehicle time-to-arrival following training in a virtual pedestrian environment

Objectives Child pedestrian injuries represent a significant public health challenge. Understanding the most complex cognitive skills required to cross streets helps us understand, improve, and protect children in traffic, as underdeveloped cognitive skill likely impacts children’s pedestrian safety. One complex component of street-crossing is the cognitive-perceptual task of judging time-to-arrival of oncoming traffic. We examined capacity of 7- and 8-year-olds to judge time-to-arrival for vehicles approaching from varying distances and speeds, as well as improvement in those judgments following intensive street-crossing training in a virtual reality (VR) pedestrian simulator. Methods 500 seven- and eight-year-olds participated in a randomized trial evaluating use of a large kiosk VR versus smartphone-based VR headset to teach street-crossing skills. Prior to randomization into VR training condition and also prior to initiation of any training, children engaged in a video-based vehicle approach estimation task to assess ability to judge traffic time-to-arrival. They then engaged in multiple VR-based pedestrian safety training sessions in their randomly assigned condition until achieving adult functioning. Soon after training and again 6 months later, children repeated the vehicle estimation task. Results Prior to randomization or training, children were more accurate judging time to arrival for closer versus farther traffic, and rapidly-moving versus slower-moving traffic, but those results were subsumed by a speed x distance interaction. The interaction suggested distance cues were used more prominently than speed cues, and speed had varying effects at different distances. Training group had minimal effect on learning and all children became significantly better at judging vehicle arrival times following training. Conclusions Children tend to underestimate vehicle arrival times. Distance cues are more impactful on time-to-arrival judgments than speed cues, but children’s estimations based both on manipulations of vehicle speed and manipulations of vehicle distance improved post-training. Improvements were retained six months later. This finding is consistent with psychophysics research suggesting vehicle approach judgments rely on optical size and looming, which are impacted both by vehicle speeds and distances. Implementation of VR-based training for child pedestrian safety is recommended, as it may improve children’s judgment of vehicle time-to-arrival, but it must be conducted cautiously to avoid iatrogenic effects.

A distributed simulation study to investigate pedestrians’ road-crossing decisions and head movements in response to different vehicle kinematics in mixed traffic

The impending deployment of automated vehicles (AVs) will lead to mixed traffic conditions, where pedestrians will be required to interact with both AVs and human-driven vehicles. For traffic flow to be safe and efficient, AVs’ understanding of pedestrians’ behaviour and intention is as important as pedestrians’ perception of AVs’ status and intent. To investigate pedestrians’ road-crossing decisions and interactive behaviour in mixed traffic, a distributed simulation study was developed by linking a CAVE-based pedestrian simulator and a desktop driving simulator. Twenty-five pairs of pedestrians and drivers were recruited, and each pair experienced 32 trials, where pedestrians decided to cross (or not) before an approaching vehicle at an un-signalised, single-lane, road. The driving pattern of the approaching vehicle (controlled by either a predefined program or human driver) and braking mode (braking/non-braking) were manipulated. For the predefined vehicles, the braking pattern was subdivided into hard braking and soft braking to provide more kinematic variability. Human drivers were also instructed to yield, or not, in different trials. Pedestrians’ road-crossing decisions and head movements were recorded and analysed. Results revealed a significant difference in head-turning patterns between crossing and non-crossing manoeuvres, demonstrating pedestrians’ head movements as a valid indicator of their road-crossing intentions. Moreover, results identified a ‘last moment check’ behaviour before pedestrians’ crossing initiation, with a significant increase in head-turning during the last 2 s. Finally, pedestrians made a similar percentage of road crossings and displayed a similar pattern of head movements, in response to human-driven and predefined vehicles, suggesting that the difference of implicit cues in the current mixed traffic setup does not impact their behaviour prior to road crossings. The findings from this study extend our knowledge of how pedestrians behave when crossing the road in mixed traffic, particularly in terms of their head-turning behaviour. We hope this information can be used by future AVs to better predict pedestrians’ road-crossing intentions in urban settings.

Interpreting pedestrians' head movements when encountering automated vehicles at a virtual crossroad

In the future, Automated Vehicles (AVs) may be able to use pedestrians’ head movement patterns to understand their crossing intentions. This ability of the AV to predict pedestrian crossing intention will improve road safety in mixed traffic situations and may also enhance traffic flow, allowing the vehicle to gradually reduce its speed in advance of a yield, eliminating the need for a complete and erratic halt. To date, most of the work conducted on studying pedestrian head movements has been based on observation studies. To further our understanding in this area, this study examined pedestrians' head movements when interacting with AVs during a range of road crossing scenarios, developed in a VR environment. Thirty-eight participants took part in this CAVE-based pedestrian simulator study. Head movements were recorded using stereoscopic motion-tracking glasses, as pedestrians crossed the road in response to an AV which approached from the right (UK-based road). A zebra crossing was included in half of the trials to understand how it affected crossing behaviour. The effect of different approaching speeds of the AV, and the presence of an external Human-Machine Interface (eHMI), on head movements and crossing behaviour was also studied. Results showed that the absolute head-turning rate (change in pedestrians' head-turning angle, per frame) increased significantly at around 1 s before a crossing initiation, reaching a peak at the crossing initiation, where pedestrians presented a “last-second check” before the crossing decision. Another increase in absolute head-turning rate to the right was seen at the end of the crossing (∼1.5 s after crossing initiation), to check the proximity of the approaching vehicle. A higher rate of head-turning was also seen for AV-non-yielding scenarios. Finally, the least number of head turns was seen for the yielding conditions which included an eHMI, in the presence of the zebra crossing. These results show the value of infrastructural and vehicle-based cues in assisting pedestrians’ crossing decisions and provide an insight into how head-turning behaviour can be used by AVs to better predict pedestrians’ crossing intentions in urban settings.

Addressing equifinality in agent-based modeling: a sequential parameter space search method based on sensitivity analysis

This study addresses the challenge of equifinality in agent-based modeling (ABM) by introducing a novel sequential calibration approach. Equifinality arises when multiple models equally fit observed data, risking the selection of an inaccurate model. In the context of ABM, such a situation might arise due to limitations in data, such as aggregating observations into coarse spatial units. It can lead to situations where successfully calibrated model parameters may still result in reliability issues due to uncertainties in accurately calibrating the inner mechanisms. To tackle this, we propose a method that sequentially calibrates model parameters using diverse outcomes from multiple datasets. The method aims to identify optimal parameter combinations while mitigating computational intensity. We validate our approach through indoor pedestrian movement simulation, utilizing three distinct outcomes: (1) the count of grid cells crossed by individuals, (2) the number of people in each grid cell over time (fine grid) and (3) the number of people in each grid cell over time (coarse grid). As a result, the optimal calibrated parameter combinations were selected based on high test accuracy to avoid overfitting. This method addresses equifinality while reducing computational intensity of parameter calibration for spatially explicit models, as well as ABM in general.

Coupling an agent-based model and an ensemble Kalman filter for real-time crowd modelling.

Agent-based modelling has emerged as a powerful tool for modelling systems that are driven by discrete, heterogeneous individuals and has proven particularly popular in the realm of pedestrian simulation. However, real-time agent-based simulations face the challenge that they will diverge from the real system over time. This paper addresses this challenge by integrating the ensemble Kalman filter (EnKF) with an agent-based crowd model to enhance its accuracy in real time. Using the example of Grand Central Station in New York, we demonstrate how our approach can update the state of an agent-based model in real time, aligning it with the evolution of the actual system. The findings reveal that the EnKF can substantially improve the accuracy of agent-based pedestrian simulations by assimilating data as they evolve. This approach not only offers efficiency advantages over existing methods but also presents a more realistic representation of a complex environment than most previous attempts. The potential applications of this method span the management of public spaces under 'normality' to exceptional circumstances such as disaster response, marking a significant advancement for real-time agent-based modelling applications.

The multi-dimensional challenges of controlling respiratory virus transmission in indoor spaces: Insights from the linkage of a microscopic pedestrian simulation and SARS-CoV-2 transmission model.

SARS-CoV-2 transmission in indoor spaces, where most infection events occur, depends on the types and duration of human interactions, among others. Understanding how these human behaviours interface with virus characteristics to drive pathogen transmission and dictate the outcomes of non-pharmaceutical interventions is important for the informed and safe use of indoor spaces. To better understand these complex interactions, we developed the Pedestrian Dynamics-Virus Spread model (PeDViS), an individual-based model that combines pedestrian behaviour models with virus spread models incorporating direct and indirect transmission routes. We explored the relationships between virus exposure and the duration, distance, respiratory behaviour, and environment in which interactions between infected and uninfected individuals took place and compared this to benchmark 'at risk' interactions (1.5 metres for 15 minutes). When considering aerosol transmission, individuals adhering to distancing measures may be at risk due to the buildup of airborne virus in the environment when infected individuals spend prolonged time indoors. In our restaurant case, guests seated at tables near infected individuals were at limited risk of infection but could, particularly in poorly ventilated places, experience risks that surpass that of benchmark interactions. Combining interventions that target different transmission routes can aid in accumulating impact, for instance by combining ventilation with face masks. The impact of such combined interventions depends on the relative importance of transmission routes, which is hard to disentangle and highly context dependent. This uncertainty should be considered when assessing transmission risks upon different types of human interactions in indoor spaces. We illustrated the multi-dimensionality of indoor SARS-CoV-2 transmission that emerges from the interplay of human behaviour and the spread of respiratory viruses. A modelling strategy that incorporates this in risk assessments can help inform policy makers and citizens on the safe use of indoor spaces with varying inter-human interactions.

Exhibition Space Circulation in Museums from the Perspective of Pedestrian Simulation

Contemporary studies largely concentrate on the physical aspects of architecture, yet within the sphere of design, the gap between user experience and the designer’s intention is an undeniable fact. This gap, illustrating the contrast between the spatial perception and the actual physical space, to some degree, mirrors preferences in human spatial behavior. It accentuates the complex relationship between human cognitive functions and spatial layout, underlining the critical role of spatial perception in architectural design and planning. This prompts the question of whether perceptions of internal traffic flow within buildings also suffer from spatial distortions. Focusing on museums, and by examining circulation paths and spatial features, a virtual museum model is devised. The research employs a holistic and reductionist approach (complex systems theory) to forge a link between circulation components and the spatial experience of architecture. Utilizing agent-based modeling tools for simulating pedestrian movements, it investigates how different circulation patterns and spatial relationships influence pedestrian behavior. The study proposes a museum circulation optimization strategy, grounded in quantifying spatial experience through Anylogic software analysis. This strategy is aimed at enhancing the design of internal traffic flows in future museum projects, offering fresh insights into museum design research, and probing into new possibilities for using pedestrian simulation software.

Urban scale pedestrian simulation in Kobe City center

We attempted to construct a digital twin of Kobe City center, from the aspect of pedestrian traffic simulation. Policy evaluation was conducted using traffic demand based on mobile phone population statistics provided by NTT DOCOMO, INC., CrowdWalk, an agent-based pedestrian simulator, and manually edited map obtained from Open Street Map to implement pedestrian signals. Background pedestrian population was estimated to be 6509, and 10,000 evacuees are assumed in the Shinko area. All of them are assumed to evacuate into three stations, JR Sannomiya, Hankyu Sannomiya, and Motomachi. The evacuation time was originally simulated to be 25,685 s without any aided policies. As a policy, splitting the evacuation route reduced the evacuation time into 17,780 s which is 69% of the original. In addition, removing signals reduced the time furthermore into 9550 s, 37% in total. Only removing the signal resulted 12,475 s, which is a reduction to 52% of the original.

Crossing roads in a social context: How behaviors of others shape pedestrian interaction with automated vehicles

Automated vehicles (AVs) are going to enter public roads in the near future and will inevitably encounter more than one pedestrian on roads. Very little is known about how multiple pedestrians will interact with AVs and their external human–machine interfaces (eHMIs). This study investigates the social influence of others on the pedestrian decision-making process during their initial encounters with AVs. In a virtual reality experiment, 26 male and 55 female participants in three experimental conditions made crossing decisions – either alone or next to other simulated pedestrians – in front of AVs equipped with abstract eHMIs showing yielding or non-yielding intentions. The results show that the behaviors of others significantly influenced pedestrian crossing decisions. While the appropriate behaviors of others contributed to faster crossing decision-making, misleading behaviors of others led to riskier crossing decisions and greater difficulty in understanding the eHMIs. However, the impacts from others on pedestrian crossing decisions disappeared once participants became more familiar with eHMIs. Qualitative data revealed that participants acquired knowledge about AVs and eHMIs through direct experience, observing others, and drawing from their past experiences. Additionally, there was a tendency among certain pedestrians to overly rely on and misuse eHMIs. To ensure the safe deployment of AVs, it is therefore essential to adequately educating the public about this novel technology and to optimize eHMI designs for effective communication with multiple pedestrians.

Enhanced Crowd Dynamics Simulation with Deep Learning and Improved Social Force Model

The traditional social force model (SFM) in crowd simulation experiences difficulty coping with the complexity of the crowd, limited by singular physical formulas and parameters. Recent attempts to combine deep learning with these models focus more on simulating specific states of crowds. This paper introduces an advanced deep social force model, influenced by crowd states. It utilizes deep neural networks to accurately fit crowd trajectory features, enhancing behavior simulation capabilities. Geometrical constraints within the model provide control over varied crowd behaviors, adjustable to simulate different crowd types. Before training, we use the SFM to refine behaviors in real trajectories with excessively small distances, aiming to enhance the general applicability of the model. Comparative experiments affirm the effectiveness of the model, showing comparable performance to both classic physical models and modern learning-based hybrid models in pedestrian simulations, with reduced collisions. In addition, the model has a certain ability to simulate crowds with high density and diverse behaviors.