Inferring vehicle spacing in urban traffic from trajectory data

This study is published in the journal "Transportation Research Part C: Emerging Technologies" with gold open access, available at https://doi.org/10.1016/j.trc.2023.104289.

Highlights

• A method to infer average 2D vehicle spacing from trajectory data is proposed.
• A perspective on the relative movement between interacting vehicles is taken.
• Empirical relations between average 2D spacing and relative speeds are identified.
• The empirical relations are termed as interaction Fundamental Diagrams (iFDs).
• iFDs describe the variation in required space for vehicle interactions.

In order to reuse the coordinate transformation into a relative movement system

Please refer to coortrans.py in the GSSM repository for a ready-to-use class.

In order to repeat the experiments:

Dependencies

tqdm, numpy, pandas, scipy, pyproj=3.2.0, joblib, matplotlib, shapely, scikit-learn

Data

Raw data can be downloaded from https://open-traffic.epfl.ch/index.php/downloads/ and saved in the folder ./RawDatasets/.

Resulting data, i.e., a zipped file of the ./OutputData/ folder, can be downloaded from https://doi.org/10.4121/8cadc255-5fd8-46ab-893a-64b76ca7b7f9.

Step-by-step instructions

Step 1. Run ./Code/Preprocessing.py to preprocess the rawdata.

Step 2. Use ./Code/IntersectionDetection.py and ./Code/IntersectionData.ipynb to identify and select intersections in the pNEUMA dataset.

Step 3. Run ./Code/Sampling_exp1-2.py, ./Code/Sampling_exp3.ipynb, and ./Code/Sampling_exp4.py to transform coordinates and sample vehicle pairs for different experiments.

Step 4. Run ./Code/Experiments.py to repeat our experiments in the article.

* ./Code/DriverSpaceInference.py is the library including classes and functions for the experiments

* We run the experiments in Linux with a cluster of CPUs. To be run on other OSs may need adjustments regarding the number of cores for parallel processing.

In order to apply the method to another dataset:

Step 1. Save raw data in the folder ./RawDatasets/.

Step 2. Create code to align the format of the new dataset to the format of the data to be saved in the folder ./InputData/.

Step 3. Design your application according to the code in ./Code/Experiments.py.

Citation

@article{Jiao2023,
  doi = {10.1016/j.trc.2023.104289},
  year = {2023},
  volume = {155},
  pages = {104289},
  author = {Yiru Jiao and Simeon C. Calvert and Sander {van Cranenburgh} and Hans {van Lint}},
  title = {Inferring vehicle spacing in urban traffic from trajectory data},
  journal = {Transportation Research Part C: Emerging Technologies}
}