SPH-planning

Swarm Navigation Framework Based on Smoothed Particle Hydrodynamics

If you are interested in swarm navigation in cluttered environments, this repository provides an SPH-based navigation framework for robot swarms. The core of our work is presented in the paper:

Swarm Navigation Based on Smoothed Particle Hydrodynamics in Complex Obstacle Environments by Ruocheng Li, Bin Xin, Shuai Zhang, Mingzhe Lyu, and Jinqiang Cui published in IEEE Robotics and Automation Letters, 2025.

If you find this work useful, please cite our paper:

@article{li2025swarm,
  title={Swarm Navigation Based on Smoothed Particle Hydrodynamics in Complex Obstacle Environments},
  author={Li, Ruocheng and Xin, Bin and Zhang, Shuai and Lyu, Mingzhe and Cui, Jinqiang},
  journal={IEEE Robotics and Automation Letters},
  year={2025},
  publisher={IEEE}
}

📺 Video demonstration: https://www.youtube.com/watch?v=4ux7pRI9-Wo

This project partially refers to the excellent implementation of Fast-Planner, and we sincerely thank the authors for their contributions.

---

🔧 Installation and Setup

Tested Environment:

• Ubuntu 20.04
• ROS Noetic

1. Install dependencies

The SPH framework requires the NLOPT optimization library. Please refer to the official documentation: 👉 https://nlopt.readthedocs.io/en/latest/NLopt_Installation/

2. Clone and compile

mkdir -p sph_planning_ws/src
cd sph_planning_ws/src
git clone https://github.com/SmartGroupSystems/SPH-planning.git
cd ..
catkin_make

3. Run the simulation

In terminal window 1:

cd sph_planning_ws
source devel/setup.bash
roslaunch water_swarm simulator.launch

Then open terminal window 2:

cd sph_planning_ws
source devel/setup.bash
roslaunch water_swarm sph.launch

At this point, an RViz simulation interface will appear. Use the 2D Nav Goal tool to click on a target position — if everything works correctly, you will observe the swarm of particles dynamically navigating toward the selected goal.

---

✈️ Real-World Experiments

In our physical experiments, we used omni-directional vision drones. If you're interested in conducting similar swarm experiments, you can check out the hardware platform we used here (not a sponsored link):

👉 https://e.tb.cn/h.h2lyknv0Acds4eq?tk=7BTfVuqPjjn

These drones are compact in size, which greatly simplifies swarm deployment and testing in practice.

Note: The real-world experiments involve many engineering tricks and low-level configurations that depend on specific hardware platforms. Moreover, instead of using ROS-native communication, we implemented a lightweight UDP-based position broadcasting system for inter-robot communication. The details of this communication protocol and its implementation will be gradually released in future updates.

---

If you have any questions, please feel free to open an issue or contact us at: 📬 ruochengli@bit.edu.cn