A mesoscopic traffic simulator designed to balance the fidelity of microscopic modeling with the scalability of macroscopic approaches.

DRIFT (DRIve Fleet Tool) represents individual vehicles as discrete agents with unique trajectories and behavioral profiles, while simultaneously capturing aggregate traffic dynamics at the network level. It provides a flexible framework for transportation researchers to study mobility dynamics, congestion patterns, and network optimization.

For detailed information about the implementation, algorithms, and theoretical background, please refer to the technical report: DRIFT - Technical Report - v1.pdf

• Individual vehicle tracking with unique IDs and behavioral profiles
• Link-level flow management for efficient large-scale simulation
• Discrete time steps with continuous-time representation
• Real-time multipliers for accelerated simulation
• Multi-day simulation horizons (up to 48 hours)

• Random Model: Uniform distribution of trips across all nodes
• Activity-Based Model: Trips based on individual activities and purposes
• Zone-Based Model: Geographic zone-based trip distribution
• Gravity Model: Distance-decay based trip generation
• Hub-and-Spoke Model: Centralized flow patterns through hub nodes

• Bureau of Public Roads (BPR) congestion function
• Dynamic capacity management
• Real-time travel time updates based on traffic volume

• Interactive network visualization with pan/zoom controls
• Color-coded agents by type and status
• Live statistics dashboard with 7 time-series plots
• Real-time trip records and performance metrics
• Network topology analysis (diameter, density, clustering)

Input Formats:

• OpenStreetMap (OSM, PBF)
• GraphML
• JSON
• Matrix Market (MTX)
• CSV edge lists

Export Capabilities:

• Trip data (CSV, JSON)
• Statistics and plots (CSV, PNG)
• Network analysis results

• Python 3.8+
• PyQt5
• NetworkX
• Matplotlib
• NumPy, SciPy
• OSMnx, GeoPandas
• Seaborn

git clone https://github.com/jbaudru/DRIFT.git
cd DRIFT
pip install -r requirements.txt
python main.py

1. Load Network: Import your network data using the "Load Graph" button

   • Supported formats: OSM, GraphML, JSON, MTX, CSV
   • Sample networks available in data/ directory

2. Configure Simulation:

   • Select source-target (s-t) model
   • Set number of agents (default: 300)
   • Adjust simulation duration and speed

3. Run Simulation: Click "Start Simulation" to begin

   • Monitor real-time visualization
   • Track statistics in the analytics dashboard
   • View trip records as they're generated

4. Export Results: Save simulation data and plots for analysis

sim-v2/
├── main.py              # Main application entry point
├── config.py            # Configuration constants
├── lib/                 # Core simulation modules
│   ├── agent.py         # Agent behavior and routing
│   ├── simulation_*.py  # Simulation engine components
│   ├── managers/        # Feature managers (UI, network, etc.)
│   └── tabs/            # UI tab implementations
├── data/                # Sample networks and simulation results
├── fig/                 # Documentation figures
└── assets/              # UI icons and resources

All nodes have equal probability for origin/destination selection.

Agents have individual activity patterns and preferences for trip generation.

Network is partitioned into geographic zones with configurable inter/intra-zone probabilities.

Trip distribution follows gravitational principles: T_ij = K * (S_i * A_j) / d_ij^β

High-centrality nodes act as hubs, concentrating flows from peripheral nodes.

• Congestion Function: BPR model with configurable α=0.15, β=4.0 parameters
• Time Resolution: Discrete time steps with continuous-time representation
• Architecture: Modular design with plugin-style s-t model integration
• Visualization: Performance-optimized rendering with level-of-detail adjustments

The project includes sample networks in data/:

• Urban networks (Brussels, Paris, Rome, etc.)
• Regional networks (Belgium, Germany, Asia)
• Test networks for algorithm validation

This project is part of ongoing research in transportation simulation. Contributions, issues, and feature requests are welcome.

This project is licensed under the MIT License - see the LICENSE file for details.

Feature	Screenshot
Network Visualization
Detailed View
Trip Records
Analytics Dashboard
Model Comparisons

Author: Julien Baudru
Contact: GitHub Profile