mlab-upenn / f1tenth_il

Implementation of various imitation learning algorithms for the F1TENTH platform

F1TENTH Imitation Learning

This repository contains code associated with A Benchmark Comparison of Imitation Learning-based Control Policies for Autonomous Racing

Quickstart

Clone this repository

git clone https://github.com/mlab-upenn/f1tenth_il.git

Navigate to the root directory of this project

cd f1tenth_il

Create a new conda environment with Python 3.8

conda create -n f110_il python=3.8

Activate the environment

conda activate f110_il

Install pip

conda install pip  

Install the dependencies for F1TENTH gym.

pip install -e .

Install other dependencies

pip install -r requirements.txt

Usage

Training

Navigate to the imitation learning folder

cd imitation_learning

Execute the training script

python train.py --algorithm=<algorithm name> --training_config=<yaml file location>

Example:

python train.py --algorithm=hg-dagger --training_config=il_config.yaml

Inference

Navigate to the imitation learning folder

cd imitation_learning

Execute the inference script

python inference.py --training_config=<yaml file location> --model_path=<model path>

Example:

python inference.py --training_config=il_config.yaml --model_path=logs/HGDAgger_model.pkl