🤖 Human-Aware Model Predictive Coverage Control (HMPCC)

This ROS package implements human-aware model predictive coverage control (HMPCC) using Model Predictive Control (MPC) combined with human motion prediction. The system coordinates a team of robots to optimally cover an environment while actively avoiding humans. This package is designed for real-time applications like search-and-rescue, smart surveillance, or service robotics in dynamic, human-populated environments.

Robots plan safe, efficient trajectories using MPC while forecasting human motion to maintain both safety and coverage performance.

---

🚀 Features

• 🧠 Human-aware, distributed multi-robot coverage control
• 📈 MPC-based real-time trajectory optimization
• 🕵️ Human motion prediction for proactive avoidance
• 🧩 ROS-native design for modular integration
• 🌍 Simulation-ready (Gazebo & RViz supported)

---

📁 Package Structure

hmpcc/
├── config/ # Configuration and tuning parameters
├── launch/ # Launch files for simulation or real robots
├── scripts/ # HMPCC evaluation scripts
├── src/ # Core algorithm implementation
├── worlds/ # Custom Gazebo worlds (optional)
├── README.md # This file

---

⚙️ Dependencies

This package was developed and tested using ROS Noetic (Ubuntu 20.04). The following dependencies are required:

ROS packages:

• geometry_msgs
• nav_msgs
• sensor_msgs
• tf2_ros
• visualization_msgs
• roscpp / rospy
• human_actor_plugin to spawn and move humans in the environment.
• turtlebot3_simulations to use TurtleBot3 robots.

External libraries:

• CasADi for solving nonlinear MPC problems
• scipy, numpy, matplotlib for human prediction modules
• Gazebo and RViz for simulation and visualization

📦 Installation

# Clone into your catkin workspace
cd ~/catkin_ws/src
git clone https://github.com/ARSControl/hmpcc.git

# Build the workspace
cd ~/catkin_ws
catkin build

# Source the workspace
source devel/setup.bash

🧠 How It Works

Each robot in the team executes the following pipeline:

1. Receive human position estimates from sensors or simulation.

2. Predict future human trajectories using motion models.

3. Solve an MPC problem to generate a safe, efficient path:

4. Maximize coverage utility

5. Avoid predicted human locations

6. Respect robot dynamics and collision constraints

7. Execute the trajectory and repeat at each control cycle.

The system supports centralized or decentralized coordination, depending on communication constraints.

📡 Inputs

• Robot positions (e.g., from tf or odometry)

• Human positions (from simulation or perception system)

• Environment map (optional)

• Robot configuration (sensor range, dynamics, etc.)

🎯 Outputs

• Safe, optimized control commands for each robot

• Trajectory visualizations in RViz

• Optionally, logs for evaluation or offline analysis

🧪 Running the Simulation

Launch the default multi-robot coverage simulation with humans:

roslaunch hmpcc sim_bringup.launch
roslaunch hmpcc multirobot_control.launch

This will:

• Spawn robots and human agents in a Gazebo world

• Launch the coverage control MPC node for each robot

• Start RViz for visualization

Launch Arguments

Argument	Description	Default
world_name	Gazebo world file	myworld.world
v_max	Max linear velocity	0.3
w_max	Max angular velocity	0.5

📈 Example Results

Below are example behaviors observed in simulation:

• Robots spread out to cover high-priority areas.

• When humans enter the field, robots modify their paths to avoid collisions.

• Smooth, real-time replanning observed under dynamic human motion.

📄 Publications & References

This package is based on concepts from:

@article{catellani2025hmpcc,
  title={HMPCC: Human-Aware Model Predictive Coverage Control},
  author={Mattia Catellani, Marta Gabbi, Lorenzo Sabattini},
  journal={IEEE International Symposium on Multi-Robot & Multi-Agent Systems (MRS)},
  year={2025}
}

🤝 Contributing

Contributions are welcome! To contribute:

1. Fork this repository

2. Create a new feature branch (git checkout -b feature-name)

3. Commit your changes

4. Push and create a pull request

Please open an issue for any bugs or feature requests.

📬 Contact

Author: Mattia Catellani

Email: mattia.catellani@unimore.it

Website: https://www.arscontrol.unimore.it/mattia-catellani/

📜 License

This project is licensed under the MIT License. See the LICENSE file for more details.