EPSILON

About

This is the project page of the paper "EPSILON: An Efficient Planning System for Automated Vehicles in Highly Interactive Environments". In this repo, we provide a simple and lightweight multi-agent simulator based on ROS and a demo implementation of the proposed EPSILON planning system.

If you use EPSILON for your academic research, please consider citing the follow

• Ding, Wenchao, et al. "EPSILON: An Efficient Planning System for Automated Vehicles in Highly Interactive Environments." IEEE Transactions on Robotics (2021).

Paper: IEEE Xplore, arXiv

Demo video: YouTube

BibTex

@article{ding2021epsilon,
  title={EPSILON: An Efficient Planning System for Automated Vehicles in Highly Interactive Environments},
  author={Ding, Wenchao and Zhang, Lu and Chen, Jing and Shen, Shaojie},
  journal={IEEE Transactions on Robotics},
  year={2021},
  publisher={IEEE}
}

The following papers are also related:

• Ding, Wenchao, et al. "Safe trajectory generation for complex urban environments using spatio-temporal semantic corridor." IEEE Robotics and Automation Letters 4.3 (2019): 2997-3004. (arXiv link)
• Zhang, Lu, et al. "Efficient uncertainty-aware decision-making for automated driving using guided branching." 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020. (arXiv link)

If you have any question, please feel free to contact us via lzhangbz@connect.ust.hk (Lu Zhang) and wdingae@connect.ust.hk (Wenchao Ding).

Prerequisites

This project has been tested on Ubuntu 16.04 (ROS Kinetic) and 18.04 (ROS Melodic). For ROS installation, please refer to the official website.

Denpendencies

• Install required packages

sudo apt-get install libgoogle-glog-dev libdw-dev libopenblas-dev gfortran

pip install empy pygame

Install OOQP

We use OOQP for solving quadratic programming problems. Please refer to link_1 and link_2 for the installation instruction.

Install Protobuf

We use Protocol Buffers for parameter configuration. For the installation guide, please refer to this link.

Build on ROS

We recommend the users create an empty workspace. Clone the repo and build:

  cd ${YOUR_WORKSPACE_PATH}/src
  git clone https://github.com/HKUST-Aerial-Robotics/EPSILON.git
  cd ..
  catkin_make
  source ~/${YOUR_WORKSPACE_PATH}/devel/setup.bash

Just have a try!

1. Open a new terminal window and start roscore:

roscore

2. Launch RViz with .rviz file:

roscd phy_simulator/rviz/
rviz -d phy_simulator_planning.rviz

3. Launch the planner's node and AI nodes:

roslaunch planning_integrated test_ssc_with_eudm_ros.launch
roslaunch ai_agent_planner onlane_ai_agent.launch

4. Launch the simulator:

roslaunch phy_simulator phy_simulator_planning.launch

Note that the simulator should be launched last.

5. We provide a simple interface for controlling the behavior of the agents:

roscd aux_tools/src/
python terminal_server.py

You can select the target agent by clicking on the colored dots and change its behavior using W-A-S-D buttons.

Acknowledgements

We would like to express sincere thanks to the authors of the following tools and packages:

• Lock-free queue: moodycamel
• Json parser: JSON for modern C++
• KD-Tree library: nanoflann
• 2D ray-casting: roguelike_FOV
• Quadratic programming: OOQP
• Cubic spline fitting: tk_spline

Licence

The source code is released under MIT license.

Disclaimer

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.