PedGrid is a minimal simulator for autonomous vehicle driving with a visualization tool. It's perfect for initial research method development with low computing resources and minimal learning curve. A team of high school and college students is creating challenging reinforcement learning tasks using OpenAI Gymnasium Framework. Currently, we are creating several indoor and outdoor environments for a variety of tasks related to human behavior and motion. Current environments:
- A corridor environment to model bidirectional pedestrian flow.
- A two-lane environment to model pedestrian crossing.
Please cite our accepted paper if you use PedGrid or doing relevant research (DOI pending):
@inproceedings{inproceedings,
author = {Muktadir, Golam Md and Huang, Taorui and Ikram, Zarif and Jawad, Abdul and Whitehead, Jim},
booktitle = {26th IEEE International Conference on Intelligent Transportation Systems ITSC 2023 (Bilbao, Bizkaia, Spain)}
year = {2023},
title = {PedGrid - A Simple yet Expressive Simulation Environment for Pedestrian Behavior Modeling}
}
We have a seperate website for documentation and tutorials here.
- It's open source
- Easy to learn: we can get started with research in a week.
- Easy to setup: written in python using python packages only.
- Grid-based: math is easier to handle.
- Easy to get metrics: We have a set of commonly used metrics in research. Data is ready for your further analysis.
conda config --append channels conda-forge
conda create -n pedgrid python=3.18.13
conda activate pedgrid
conda install gym=0.21
conda install matplotlib
python manual_control.py --env Pedestrian-Empty-9x16-v0
python manual_control.py --env MultiPedestrian-Empty-9x16-v0
