Bi-Level Optimization Augmented with Conditional Variational Autoencoder for Autonomous Driving in Dense Traffic

This repository contains the source code to reproduce the experiments in our IEEE CASE 2023 paper Bi-Level Optimization Augmented with Conditional Variational Autoencoder for Autonomous Driving in Dense Traffic.

Getting Started

1. Clone this repository:

git clone https://github.com/jatan12/MPC-Bi-Level.git
cd MPC-Bi-Level

2. Create a conda environment and install the dependencies:

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

3. Download CVAE Initialization Models and extract the zip file to the weights directory.

Reproducing our main experimental results

MPC-Bi-Level

python main_bilevel.py --density ${select} --four_lane ${True / False for two lane}

MPC Baselines

To run a baseline {vanilla, grid, random, batch}:

python main_baseline.py --baseline ${select} --density ${select} --four_lane ${True / False for two lane}

Note: Default number of episodes is 50. To record / render the environment:

python main_baseline.py --episodes ${select} --record True --render True

Learning Good Initialization Distribution

1. Clone the Deep Declarative Networks repository:

cd MPC-Bi-Level
git clone https://github.com/anucvml/ddn.git

2. Download the training dataset and extract the zip file to the dataset directory.

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