Multi-Robot Warehouse Orchestration

A complete multi-robot autonomous warehouse system built on ROS2 Humble. Four AGV robots autonomously pick up and deliver items in a Gazebo-simulated warehouse, coordinated by a central Fleet Manager with conflict-free path planning.

Architecture

┌─────────────────────────────────────────────────────┐
│                    Fleet Manager                    │
│  ┌──────────────┐  ┌──────────────┐  ┌───────────┐  │
│  │Task Allocator│  │ CBS Resolver │  │ State Mgr │  │
│  │ (Hungarian / │  │ (Conflict -  │  │ (Battery, │  │
│  │  Greedy)     │  │ Based Search │  │ Heartbeat)│  │
│  └──────────────┘  └──────────────┘  └───────────┘  │
└──────────────────────────┬──────────────────────────┘
                           │ Nav2 Actions
        ┌────────────┬─────┼──────┬────────────┐
        ▼            ▼            ▼            ▼
   ┌─────────┐  ┌─────────┐  ┌─────────┐  ┌─────────┐
   │ Robot 1 │  │ Robot 2 │  │ Robot 3 │  │ Robot 4 │
   │ Nav2+   │  │ Nav2+   │  │ Nav2+   │  │ Nav2+   │
   │ SLAM    │  │ SLAM    │  │ SLAM    │  │ SLAM    │
   └────┬────┘  └────┬────┘  └────┬────┘  └────┬────┘
        └────────────┴────────────┴────────────┘
                           ▼
                  ┌──────────────────┐
                  │  Gazebo Sim      │
                  │  (Warehouse +    │
                  │   AGV robots)    │
                  └──────────────────┘
                           │
                  ┌──────────────────┐
                  │  Web Dashboard   │
                  │  Flask+SocketIO  │
                  │  localhost:5000  │
                  └──────────────────┘

Package Structure

Package	Description
warehouse_msgs	Custom ROS2 msgs, srvs, actions
warehouse_description	AGV URDF/Xacro, Gazebo world
warehouse_gazebo	Simulation launch files
warehouse_navigation	Nav2 + SLAM config
fleet_manager	Task allocation, CBS, state management
warehouse_dashboard	Flask+SocketIO web UI

Quick Start

Prerequisites

• ROS2 Humble
• Gazebo Classic 11
• ros-humble-navigation2, ros-humble-slam-toolbox

# Install ROS2 dependencies
cd ~/multi_robot_warehouse_ws
rosdep install --from-paths src --ignore-src -r -y

# Install Python deps
pip3 install flask flask-socketio eventlet scipy numpy

# Build
colcon build --symlink-install
source install/setup.bash

Launch Full Simulation

# Full system (Gazebo + 4 robots + Nav2 + Fleet Manager + Dashboard)
ros2 launch warehouse_gazebo full_simulation.launch.py

# Dashboard available at: http://localhost:5000

Step-by-Step Launch

# 1. Start Gazebo world
ros2 launch warehouse_gazebo warehouse_world.launch.py

# 2. Spawn all robots (in new terminal)
ros2 launch warehouse_gazebo spawn_multi_robot.launch.py

# 3. Start Nav2 (requires warehouse_map.yaml)
ros2 launch warehouse_navigation multi_nav2_bringup.launch.py

# 4. Start Fleet Manager
ros2 launch fleet_manager fleet_manager.launch.py

# 5. Start Dashboard
ros2 launch warehouse_dashboard dashboard.launch.py

Building the Map (Faz 3)

# Start SLAM on robot_1 and teleoperate to map the warehouse
ros2 launch warehouse_navigation slam.launch.py robot_name:=robot_1

# Drive robot_1 with keyboard
ros2 run teleop_twist_keyboard teleop_twist_keyboard --ros-args -r cmd_vel:=/robot_1/cmd_vel

# Save map when done
ros2 run nav2_map_server map_saver_cli -f src/warehouse_navigation/maps/warehouse_map

Docker

cd docker
docker-compose up --build
# Dashboard: http://localhost:5000

Sending Tasks Manually

# Via ROS2 topic
ros2 topic pub /task_requests warehouse_msgs/msg/TaskRequest \
  '{task_id: "test_1", task_type: 0, pickup_zone: "loading_dock_1",
    dropoff_zone: "unloading_dock_1", priority: 1,
    pickup_location: {position: {x: -14.0, y: 3.0, z: 0.0}},
    dropoff_location: {position: {x: 14.0, y: 3.0, z: 0.0}}}'

# Via service
ros2 service call /assign_task warehouse_msgs/srv/AssignTask \
  '{task: {task_id: "test_2", task_type: 0, priority: 2,
    pickup_location: {position: {x: -14.0, y: -3.0, z: 0.0}},
    dropoff_location: {position: {x: 14.0, y: -3.0, z: 0.0}}}}'

# Via dashboard: http://localhost:5000

Fleet Manager Algorithms

Task Allocation

• Greedy: O(n) — assigns each task to closest battery-sufficient robot
• Hungarian: O(n³) — globally optimal batch assignment

Conflict Resolution (CBS)

1. Plan individual A* paths for each robot
2. Detect spatiotemporal vertex conflicts
3. Branch constraint tree (constrain robot A or B at conflict point)
4. Replan with constraints until no conflicts remain
5. Fallback: priority-based planning if CBS exceeds timeout

Battery Management

• Moving: ~0.5%/s drain → ~3.3 min runtime
• Charging: ~2%/s → ~50s to full
• Auto-dispatch to nearest free charging station at 20%

Warehouse Layout

W=32m, H=22m
    Loading Docks          Shelves (4×5)         Unloading Docks
    LD1 (-14,3)    R1  R2  R3  R4                UD1 (14,3)
    LD2 (-14,-3)   rows at x=-10,-4,4,10         UD2 (14,-3)

    Charging Stations (corners):
    CS1(-13,9)  CS2(13,9)  CS3(-13,-9)  CS4(13,-9)

Topics Reference

/fleet_status              warehouse_msgs/FleetStatus  @ 1Hz
/task_requests             warehouse_msgs/TaskRequest
/task_results              warehouse_msgs/TaskResult
/robot_X/scan              sensor_msgs/LaserScan
/robot_X/odom              nav_msgs/Odometry
/robot_X/cmd_vel           geometry_msgs/Twist

Services:
/assign_task               warehouse_msgs/srv/AssignTask
/cancel_task               warehouse_msgs/srv/CancelTask
/get_robot_status          warehouse_msgs/srv/GetRobotStatus

Komutlar ve kullanım kılavuzu için: COMMANDS.md