Decentralized Multi-Robot Belief Space Planning (Turtlesim + ROS)

This project demonstrates decentralized coordination between two robots (simulated turtles) to reduce uncertainty in a shared belief map through local movement and inter-robot communication. It is implemented in ROS Noetic using Turtlesim and simulates coordination through a belief-sharing protocol.

---

🐢 Project Features

• Two robots independently estimate a 5x5 belief grid
• Each robot reduces uncertainty (belief) based on physical location
• Robots predict each other’s behavior to avoid redundant work
• When predictions differ, they share beliefs and update accordingly
• Visual distinction via colored pen trails (red for R1, green for R2)

---

📁 Project Structure

multi_robot_bsp/
├── launch/
│   └── multi_robot.launch
├── msg/
│   └── BeliefGrid.msg
├── scripts/
│   ├── robot1_node.py
│   └── robot2_node.py
├── CMakeLists.txt
└── package.xml

---

🚀 How to Run

⚠️ Assumes ROS Noetic and catkin_ws already set up.

1. Build the workspace

cd ~/catkin_ws
catkin_make
source devel/setup.bash

2. Launch the system

roslaunch multi_robot_bsp multi_robot.launch

This will:

• Start two turtlesim windows
• Assign red and green pens
• Launch control nodes for both robots

---

💡 How It Works

Each robot:

• Maintains a local 5x5 belief grid
• Reduces uncertainty by visiting grid cells
• Predicts the other robot’s movement
• If predictions don’t match, triggers communication
• Publishes its belief grid over a shared topic
• Updates its own belief using received data

---

📦 Custom Message

The BeliefGrid.msg format:

float32[] grid     # 25 floats (row-major order)
string sender_id   # "robot1" or "robot2"

---

🎯 Goals

• Demonstrate decentralized planning under partial information
• Simulate robot coordination without centralized control
• Test basic belief-space planning ideas with ROS

---

📈 System Architecture Diagram

     +---------------------+            +---------------------+
     |     Robot 1 (R1)    |            |     Robot 2 (R2)    |
     |---------------------|            |---------------------|
     |  Local Belief Grid  |            |  Local Belief Grid  |
     |  + Motion Planner   |            |  + Motion Planner   |
     |  + Target Predictor |            |  + Target Predictor |
     +---------|-----------+            +----------|----------+
               |                                   |
               |       Predicts Other's Target     |
               |<------------------------------->  |
               |                                   |
     +---------v-----------+            +----------v----------+
     | If Disagreement,    |            |  If Disagreement,   |
     | Send BeliefGrid.msg |            |  Send BeliefGrid.msg|
     +---------|-----------+            +----------|----------+
               |                                   |
               |      /belief_channel (topic)      |
               +-------------^---------------------+
                             |
                    +------------------+
                    | BeliefGrid.msg   |
                    |  - grid (25)     |
                    |  - sender_id     |
                    +------------------+

Turtlesim UI:

• R1 draws RED trail
• R2 draws GREEN trail

Movement:

• R1 seeks max uncertainty
• R2 seeks min uncertainty
• Grid cell = 2 + i * 2 (in 1–10 range)