MEDIRL-IC

Article

Enhancing Pedestrian Route Choice Models through Maximum-Entropy Deep Inverse Reinforcement Learning with Individual Covariates (MEDIRL-IC): This article has been accepted for publication in IEEE Transactions on Intelligent Transportation Systems. You can access it via IEEE Xplore.

Introduction

This project is a collection of algorithms and models dedicated to "Deep Inverse Reinforcement Learning with Individual Covariates" in the context of pedestrian route choice. Developed by Boyang Li at SUPD, Peking University.

Directory Structure

• A_star: Implementation of the A* algorithm.
• DTW: Initial setup and configurations for DTW (Dynamic Time Warping).
• data: Contains datasets, data preprocessing scripts, and other data-related utilities.
• model: Stores model parameters and configuration files.
• plot: Scripts for generating visualizations and related plots.
• realGrid: Urban grid classes and methods representing real-world scenarios.
• img: Images utilized primarily for the project's README documentation.

Configuration and Utility Files

• .gitignore: Configuration file for Git to determine which files and directories to ignore before committing.
• README.md: Provides an overview and documentation for the project.
• img_utils.py: Utility functions related to image processing.
• tf_utils.py: Utility functions related to TensorFlow operations.
• utils.py: General utility functions for the project.

Causal Analysis

• causal_data.py: Script for handling causal data.
• causal_learn_pc_detailed.py: Detailed learning scripts for the PC algorithm in causal inference.
• causal_plot_detailed.ipynb: Jupyter notebook for detailed plotting and visualization of causal data.

Deep Inverse Reinforcement Learning (IRL)

• deep_irl_be.py: Deep IRL only considered built environment.
• deep_irl_realworld.py: Implementation of deep IRL with IC for real-world scenarios.
• demo_deepirl_be.py: Demonstration script for deep IRL backend.
• demo_deepirl_realworld.py: Demonstration script for deep IRL with IC in real-world scenarios.

Recursive Logit Model

• demo_recursive_logit.py: Demonstration script for the recursive logit model.
• recursive_logit.py: Implementation of the recursive logit model.
• test_recursive_logit.py: Testing script for the recursive logit model.

DNN-PSL Model

• dnn_psl.py: Script related to the deep neural network model.
• test_dnn_psl.py: Testing script for the DNN model.

Trajectory Evaluation and Analysis

• evaluate_traj_IRL.py: Evaluation scripts for IRL trajectories.
• evaluate_traj_psl.py: Evaluation scripts for PSL trajectories.
• traj_contrast.py: Scripts for trajectory contrast analysis.
• traj_contrast_psl.py: Scripts for PSL trajectory contrast analysis.
• traj_policy_logll.py: Scripts related to policy log likelihood for trajectories.
• trajectory.ipynb: Jupyter notebook for trajectory functions.
• trajectory.py: Scripts related to trajectory functions.

Miscellaneous

• nshortest_path.ipynb: Jupyter notebook related to the choice set of path-size logit model.

Getting Started

1. Clone the repository.
2. Install necessary dependencies.
3. Run the desired scripts or models.

Dependencies

• Fiona: 1.8.13
• GDAL: 3.0.4
• geopandas: 0.8.2
• matplotlib: 3.0.3
• networkx: 2.4
• numpy: 1.14.5+mkl
• pandas: 0.25.3
• pyproj: 2.5.0
• Rtree: 0.9.3
• scipy: 1.4.1
• seaborn: 0.9.1
• Shapely: 1.6.4.post2
• tensorflow: 0.12.1

Data Availability

Due to individual privacy concerns, we only provide geographical data for the training region along with a limited set of encrypted individual trajectory data.

Results and Visualizations

DeepIRL

Box Comparison

Reward and Likelihood Plot

Dynamic Traffic Equilibrium

Reward Map based on IC

SZ Reward Map

Causal Directed Graph

Causal Strength

License and Credits

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

Citation

Li, B., & Zhang, W. (2024). Enhancing Pedestrian Route Choice Models Through Maximum-Entropy Deep Inverse Reinforcement Learning With Individual Covariates (MEDIRL-IC). IEEE Transactions on Intelligent Transportation Systems, 1-18. https://doi.org/10.1109/TITS.2024.3457680

@ARTICLE{10689250,
  author={Li, Boyang and Zhang, Wenjia},
  journal={IEEE Transactions on Intelligent Transportation Systems}, 
  title={Enhancing Pedestrian Route Choice Models Through Maximum-Entropy Deep Inverse Reinforcement Learning With Individual Covariates (MEDIRL-IC)}, 
  year={2024},
  pages={1-18},
  keywords={Pedestrians; Analytical models; Decision making; Predictive models; Reinforcement learning; Biological system modeling; Trajectory; Deep inverse reinforcement learning; pedestrian route choice; causal discovery; cell phone signaling},
  doi={10.1109/TITS.2024.3457680}
}