wingsweihua / imingail

ImInGAIL: Learning to Simulate on Sparse Trajectory Data (ECML-PKDD'20)

@inproceedings{imingail,
 author = {Wei, Hua and Chen, chacha and Liu, Chang and Zheng, Guanjie and Li, Zhenhui},
 title = {Learning to Simulate on Sparse Trajectory Data},
 booktitle = {Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
 series = {ECML-PKDD '20},
 year = {2020},
 organization={Springer}
} 

Usage and more information can be found below.

Usage

How to run the code:

The code relies on the simulator of CityFlow, which provides scalable, multi-process simulation for transportation sceanrios.

1. Intall the gym environment 'gym_citycar-v0' for CityFlow in your python environment.

   https://github.com/wingsweihua/gym-cityflow-car.git

   There are two branches, one for intersection network and one for ring network.

2. Pull the codes for ImInGAIL.

   git clone https://github.com/wingsweihua/imingail.git

Start an experiment in the python environment with 'gym_citycar-v0' with the following script:

``bash runexp_gail_sparse.sh``

Dataset

• synthetic data

  • simulation file: Traffic file and road networks can be found in data/1_1 && data/ring.
  • logged data: data/expert_trajs/1_1 && data/expert_trajs/ring

• real-world data

  • simulation file: Traffic file and road networks can be found in data/1x4_LA && data/4x4_gudang.
  • logged data: data/expert_trajs/1_4 && data/expert_trajs/4_4