End-to-End Learning of Behavioural Inputs for Autonomous Driving in Dense Traffic

This repository contains the source code to reproduce the experiments in our IROS 2023 paper. Video of results

Getting Started

1. Clone this repository:

git clone https://github.com/jatan12/DiffProj.git
cd DiffProj

2. Create a conda environment and install the dependencies:

conda create -n diffproj python=3.8
conda activate diffproj
pip install -r requirements.txt

3. Download Trained Models to the weights directory.

Reproducing our main experimental results

Ours

Four Lane

python main_diffproj.py --density ${select} --render True

Two Lane

python main_diffproj.py --density ${select} --two_lane True --render True

Baselines

To run a baseline {batch, grid, mppi}:

Four Lane

python main_baseline.py --baseline ${select} --density ${select} --render True

Two Lane

python main_baseline.py --baseline ${select} --density ${select} --two_lane True --render True

Training the Behavioral Input Distribution Model

1. Download the training dataset and extract the files to the dataset directory.

2. The training example is shown in the Jupyter Notebook and can also be viewed using Notebook Viewer.