romanzy313/car-following-project

IFEEMCS520100 Difference in car-following interaction between following a human-driven vehicle and an autonomous vehicle project

Human Drivers' Reactions to Automated Vehicles: An Analysis on Traffic System Impact

Source code for final project of IFEEMCS520100 Fundamentals of Artificial Intelligence Programme course at TU Delft.

Authors

• Roman Volovoy ([email protected])
• Xianzhi Zhang ([email protected])
• Yuncong Liu ([email protected])
• Zhengchu Wu ([email protected])

Getting started

Pipenv is used for dependency management, install it if needed.

pip install pipenv

Install all packages with dev dependencies from Pipfile.lock; this will take a while.

pipenv install -d

Enter python environment

pipenv shell

Run a simple example. It will run an example simulation and then launch a visualization of the result.

python example_run.py && python visualization.py --file ./results/test_run.json

Rebuilding AI training data

This repo contains trained models of all different datasets (2 HA, 2 HH, 1 AH) clusters. Re-clustering and re-training can be performed by running sh ./train.sh. This process can take a very long time (approx. 2 hours on 12 core machine)

Run tests

There are many small unit tests to verify that the models work by themselves. These tests can be run with the following command:

pytest -n 2 tests.py

Running the tests will output all simulation results in the test_results folder. The output can be viewed by running

python visualization.py -f test_results/<test-name>.json