JuPedSim App
Video Tutorial: "Get Started with JuPedSim"
Web-Based JuPedSim — Pedestrian Crowd Simulation in Your Browser
A free, browser-based interface for pedestrian crowd dynamics research and evacuation analysis
JuPedSim has long been an engine for pedestrian dynamics research. This web interface was built to remove the last remaining barrier to entry: the setup. Researchers, safety engineers, students and anyone who is interested can now access the full simulation pipeline from any device, directly in a browser tab — no installation is required.
The tool is designed for scientific work. It supports import of common file formats (such as DXF from CAD tools and IFC from BIM data) and exports to SQLite for post-analysis in Python.
Building a simulation
Drawing geometry
The canvas-based drawing tools let you sketch walkable areas, walls, obstacles, exits, and agent start areas by hand. For precise work, the interface borrows conventions from Blender — axis locking (X/Y), typed dimensions, and snap-to-origin allows you to conveniently modify the details. A background image overlay with scale calibration is useful when working from floor plans or photographs.
If you already have a model of the geometry, you can import DXF files from CAD software or IFC files from BIM. Please note that these files must comply a specific strcuture: only simple and closed polygons are permitted as gemometry elements. Once this is done you can drop the geometry files directly onto the canvas, they parse immediately — no separate import dialog needed.
Build a complete simulation from scratch: draw geometry, configure agent journeys, and run.
Axis locking, typed dimensions, snap-to-origin, and live distance/angle tooltips for accurate geometry.
Simulation models
Five pedestrian models are available, each suited to different research questions. You can run identical scenarios across all of them to isolate model-specific behavior.
Journeys and flow control
Agent behavior is defined through a journey system. A journey consists of various stages that are completed between the start and the destination. Stages can be: waypoints, waiting areas with configurable pause durations, and areas with flow limitations e.g. for modeling limited usability of doors. Stage specific parameters are visually indicated on the canvas so the logic of a scenario stays readable at a glance.
Define agent routes with stages to control flow and waiting behavior.
Analysis and results
After a simulation run, trajectories are visualized with adjustable playback speed. Density heatmaps are computed using Gaussian kernel estimation via PedPy. Speed distributions and evacuation time curves are generated automatically. For statistical robustness, batch simulations can be seeded across multiple runs.
Results export to SQLite for use in custom Python pipelines, or to a ZIP archive with agents.csv and trajectory.csv ready for further processing. You can also import existing JuPedSim results directly into the interface for visualization without re-running the simulation.
Import existing JuPedSim results for visualization, parse IFC (BIM) files, and manage project files.
Research applications
The most common use case is evacuation analysis — simulating scenarios to compute evacuation times and identify bottlenecks before they exist in the real world. Since the geometry can be imported from BIM models, this kind of analysis can happen early in the design process rather than as a post-hoc check.
Model comparison is another area where the browser interface pays off. Running identical scenarios across all five built-in models side by side makes it easy to understand how modeling assumptions affect results — a conversation that often stays abstract in the literature becomes immediately visible.
For teaching, the zero-install nature of the tool makes it practical in a course setting. Students can be working through their first simulation within minutes of receiving a link, without any local environment issues to debug first.
Researchers who need to go deeper can export to SQLite and pick up in Python using PedPy, the analysis library that also powers the in-browser density heatmaps.
Technical details
The frontend is built on React 19 with a canvas-based drawing layer and Material-UI components. The backend runs FastAPI on Python, coordinating with the JuPedSim engine, alongside an Express/Node.js layer for other services. Scenarios are stored in MongoDB on German research infrastructure. Authentication uses Helmholtz AAI for institutional access, and an integrated AI assistant runs on Helmholtz Blablador with RAG-based retrieval over the documentation.
Roadmap (Upcoming Features)
The next development phase of JuPedSim Web Community focuses on strengthening scientific reliability, AI-assisted workflows, and fire–evacuation coupling.
First, we will expand validation and verification capabilities by implementing standardized benchmark scenarios. This includes integrating RiMEA test cases and ISO 20414 test cases, and properly tagging these scenarios within the scenario database to support reproducibility and systematic evaluation of simulation results.
Second, we will transform the current AI assistant into a simulation copilot using the Model Context Protocol (MCP). The copilot will enable users to draft scenarios from natural language, automatically validate and repair scenarios before execution, run and compare simulation with experiments at scale, and generate actionable analytical insights rather than simple chat responses.
Third, we plan to introduce coupling with Fire Dynamics Simulator (FDS) outputs to support fire–evacuation analysis. This integration will allow JuPedSim agents to read FDS slice data to estimate Fractional Effective Dose (FED) for each agent and dynamically adjust walking speed based on visibility conditions. The coupling will rely on slice data representing FED-related compounds and visibility fields at a standardized height from the floor, following the approach used in FDS+Evac.
Get started
Visit app.jupedsim.org and log in with Helmholtz AAI (also via external accounts) to create and save scenarios. Draw your walkable geometry using the canvas tools, or drop a DXF or IFC file directly onto the canvas to import it. Add agent start areas, define their journeys, select a pedestrian model, and run. Results appear in the playback view immediately. When you’re ready to go further, export to SQLite or CSV and continue your analysis in Python with PedPy.
The Web-Based simulator is an experimental platform that accesses JuPedSim’s open-source models. We welcome feedback, ideas, and contributions from the community. If you have suggestions for improvements or new features, please share them. Community input helps us prioritise the roadmap and guide future development.