This is the project page of the paper "Real-Time Trajectory Planning for Autonomous Driving with Gaussian Process and Incremental Refinement", to appear in ICRA 2022. GPIR is developed aiming to provide an unified real-time trajectory planning framework in dynamic environment for autonomous driving.
Paper: ArXiv
Demo video: YouTube
Author: Jie Cheng, Yingbing Chen, Qingwen Zhang, Lu Gan and Ming Liu, RAM-LAB
If you find this repo helpful for your research, please cite our related paper.
@inproceedings{Jie2022gpir,
title = {Real-Time Trajectory Planning for Autonomous Driving with {G}aussian Process and Incremental Refinement},
author = {Cheng, Jie and Chen, Yingbing and Zhang, Qingwen and Gan, Lu and Liu, Chengju and Liu, Ming},
booktitle = {2022 IEEE International Conference on Robotics and Automation (ICRA)},
year = {2022},
organization = {IEEE}
}
Our software is developed and tested in Ubuntu20.04 with ROS noetic. Follow this link to install noetic.
We use Carla-0.9.11 as the simulator (download via this link), and recommend to use a conda environment.
conda create -n carla_py37 python=3.7
To use the Carla PythonAPI, you need to export the path or install it
export PYTHONPATH=$PYTHONPATH:/carla_directory/PythonAPI/carla/dist/carla-0.9.11-py3.7-linux-x86_64.egg
or
pip install /carla_directory/PythonAPI/carla/dist/carla-0.9.11-py3.7-linux-x86_64.egg
Note that ROS melodic or below use Python2, and you may need to change python version above accordingly.
cd ${YOUR_WORKSPACE}/src
git clone --recurse-submodules https://github.com/jchengai/gpir.git
cd gpir
conda activate carla_py37
pip install -r ./requirements.txt
bash ./setup.sh
You may check the pip path via which pip
.
cd ../../
catkin_make -DCMAKE_BUILD_TYPE=RELEASE -j4
source ./devel/setup.sh
cd to_your_carla_folder && ./carla.sh
roslaunch planning_core carla_setup.launch
roslaunch planning_core planning.launch
Since this repo does not contain the behavior planner, we provide a simple interface for manually controlling the behavior of the Ego vehicle.
roslaunch planning_core keyboard.launch
We would like to express sincere thanks to the authors of the following tools and packages:
The source code is released under MIT license.
This is research code, it is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of merchantability or fitness for a particular purpose.