Edge / fog computing architecture
The system is designed around an edge/fog computing model. The simulation (or in a real deployment, the physical robot and its sensors) runs on-site as the edge layer. The digital twin processes navigation intelligence locally or on a nearby fog server. Foxglove acts as a remote supervision panel (RP2) that can be accessed from any computer, either on the same machine, on the local network, or remotely over the cloud via an API.
This architecture means that:
- The edge (robot + environment sensors) produces raw data on-site
- The fog/cloud (digital twin) processes that data and computes navigation commands
- The RP2 (Foxglove) allows a human operator to monitor, visualise, and send commands from anywhere, all that is needed is the IP address of the machine running the digital twin
┌──────────────────────────────┐
│ Edge / On-site │
│ Gazebo (or real robot) │
│ Sensors, motors, lidar │
└──────────┬───────────────────┘
│ ROS2 topics
▼
┌──────────────────────────────┐
│ Fog / Cloud │
│ Digital twin (Nav2, AI) │
│ Foxglove bridge :8765 │
└──────────┬───────────────────┘
│ WebSocket
▼
┌──────────────────────────────┐
│ RP2 — Remote panel │
│ Foxglove (any browser) │
│ ws://<ip>:8765 │
└──────────────────────────────┘
Connection setup
- Open https://app.foxglove.dev in a browser (Chrome recommended)
- Select Open connection → Foxglove WebSocket
- Enter
ws://<ip>:8765 — use localhost if running on the same machine, or the IP address of the computer running the digital twin if accessing remotely
- The Foxglove bridge runs on port 8765, started automatically by
hospital.launch.py
Installing the layout
A pre-configured layout is saved in the project repository:
Foxglove/foxglove_layout/CloudTwin.json
To load it: in Foxglove, click Layout (top-left) → Import layout → select CloudTwin.json.
Installing the room command panel
The custom Foxglove extension for voice/text room commands is in:
Foxglove/foxglove_panels/room-command-panel/
Pre-built .foxe files are available in that directory. To install:
- Transfer the
.foxe file to the computer running Foxglove (if different from the build machine)
- In Foxglove → Profile icon → Extensions → Install local extension
- Select the
.foxe file
Please use teh latest version of the .foxe file. To this date the latets version is vasileiosfilipposskarleas.room-command-panel-2.0.0
To rebuild from source:
cd ~/Documents/CloudTwin/Foxglove/foxglove_panels/room-command-panel
npm install
npm install --save-dev @foxglove/extension
npm run package
Layout structure
The layout is split into two main areas: a 3D map panel on the left (50% width), and a tabbed panel on the right with three tabs.
┌──────────────────────┬──────────────────────┐
│ │ [Navigation] [Debug]│
│ │ [Teleop] │
│ 3D map panel ├──────────────────────┤
│ │ │
│ - Hospital map │ Tab content varies │
│ - Robot position │ (see below) │
│ - Planned path │ │
│ - Crowd density │ │
│ - People positions │ │
│ │ │
└──────────────────────┴──────────────────────┘
3D map panel (left side)
The main visualisation panel. Fixed frame is map, follow mode is follow-pose.
Visible topics
| Topic | Type | What it shows |
/map | OccupancyGrid | Static hospital map (dark blue walls, red obstacles) |
/people_positions | PoseArray | Green arrows showing current human positions |
/crowd_density | OccupancyGrid | Turquoise overlay showing Gaussian density around people |
/plan | Path | Nav2 global planned path |
/plan_smoothed | Path | Smoothed version of the global plan |
/local_plan | Path | Nav2 local planner trajectory |
/received_global_plan | Path | Path received by the local planner |
/global_costmap/costmap_updates | OccupancyGrid | Global costmap incremental updates |
/local_costmap/costmap_updates | OccupancyGrid | Local costmap updates |
/cost_cloud | PointCloud2 | DWB cost evaluation point cloud |
/robot_description | RobotModel | bcr_bot 3D model |
/initialpose | PoseWithCovarianceStamped | Initial pose estimate |
Hidden topics (available for debug)
| Topic | Why hidden |
/map_dynamic | Same as /map but with crowd walls — redundant with /crowd_density visually |
/bcr_bot/scan | Lidar scan — clutters the view |
/global_costmap/costmap | Full costmap — heavy to render, use updates instead |
/local_costmap/costmap | Same reason |
/bcr_bot/kinect_camera/* | Camera feeds shown in the Teleop tab instead |
Publishing goals
The 3D panel is configured to publish pose goals on /goal_pose_foxglove when you click on the map. The goal_relay node forwards these to Nav2's /goal_pose with corrected timestamps.
To send a goal: select the Publish pose tool in the 3D panel toolbar, then click and drag on the map to set position and orientation.
Navigation tab
The primary operational tab. Contains the room command panel and real-time telemetry plots.
┌──────────────────────────────────┐
│ Room command panel │ ← Text/voice destination commands
│ [Text input] [Send] [🎤] │
│ [Language toggle] [Feedback] │
├──────────────────────────────────┤
│ /bcr_bot/cmd_vel (linear) │ ← Linear velocity plot (x, y, z)
├──────────────────────────────────┤
│ /bcr_bot/cmd_vel (angular) │ ← Angular velocity plot (x, y, z)
├──────────────────────────────────┤
│ /bcr_bot/imu angular velocity │ ← IMU angular velocity (gyroscope)
├──────────────────────────────────┤
│ /bcr_bot/imu linear accel. │ ← IMU linear acceleration
└──────────────────────────────────┘
Room command panel
The custom extension that allows sending navigation commands:
- Type a room name (e.g., "Go to urgences", "Salle 101") and press Enter
- Click 🎤 to trigger voice recording on the microphone connected to the digital twin's computer (requires
speech_node running via logic.launch.py)
- Language toggle switches between French (default) and English for speech recognition
- Feedback area shows green for "Navigating to..." or red for "Room not found"
Telemetry plots
Four real-time plots showing:
- cmd_vel linear (x, y, z) : the velocity commands Nav2 sends to the robot
- cmd_vel angular (x, y, z) : the rotation commands
- IMU angular velocity : what the robot's gyroscope actually measures
- IMU linear acceleration : accelerometer readings (gravity visible on z-axis)
Comparing cmd_vel (commanded) vs IMU (measured) helps verify the robot is executing commands correctly.
Debug tab
For troubleshooting and manual testing.
┌──────────────────┬──────────────────┐
│ Publish panel │ Diagnostics │
│ /goal_pose_ │ summary │
│ foxglove │ │
├──────────────────┴──────────────────┤
│ ROS logs │
├──────────────────┬──────────────────┤
│ Raw messages │ Markdown │
│ /goal_pose_ │ (launch │
│ foxglove │ instructions) │
└──────────────────┴──────────────────┘
- Publish panel — manually publish a PoseStamped goal on
/goal_pose_foxglove with specific coordinates
- Diagnostics — Nav2 diagnostic messages (planner status, controller status)
- ROS logs — live log output from all running nodes (filtered to INFO and above)
- Raw messages — inspect the raw content of
/goal_pose_foxglove messages
- Markdown — quick-reference launch instructions
Teleop tab
For manual robot control and camera monitoring.
┌──────────────────┬──────────────────┐
│ Depth camera │ Teleop joystick │
│ /bcr_bot/ │ publishes to │
│ kinect_camera/ │ /bcr_bot/ │
│ depth/image_raw │ cmd_vel │
├──────────────────┼──────────────────┤
│ RGB camera │ Transform tree │
│ /bcr_bot/ │ TF hierarchy │
│ kinect_camera/ │ visualisation │
│ image_raw │ │
└──────────────────┴──────────────────┘
- Depth camera — raw depth image from the Kinect-style sensor
- RGB camera — colour camera feed from the robot
- Teleop joystick — arrow-key style controls publishing directly to
/bcr_bot/cmd_vel (bypasses Nav2 for manual driving)
- Transform tree — visual hierarchy of all TF frames (map → odom → base_footprint → wheels, sensors, etc.)
Project file structure
Foxglove/
├── foxglove_layout/
│ └── CloudTwin.json # Foxglove layout configuration
└── foxglove_panels/
└── room-command-panel/
├── src/
│ ├── index.ts # Extension entry point
│ └── RoomCommandPanel.ts # Panel UI + logic
├── package.json
└── *.foxe # Built extension files (versioned)
