Add MAPPO Device Connect dashboard Learning Path

content/learning-paths/servers-and-cloud-computing/use-mappo-device-connect-dashboard/1-prepare-dashboard.md

@@ -0,0 +1,141 @@
+---
+title: Prepare the Device Connect dashboard
+description: Install the dashboard dependencies and verify that the exported MAPPO actor matches the deployment interface.
+weight: 2
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Understand the deployment workflow
+
+The training Learning Path produces an actor-only `.npz` file. The dashboard treats that file as a model that can move through a Device Connect deployment:
+
+```text
+Exported MAPPO actor
+        ↓
+Local model server
+        ↓
+Device Connect model download
+        ↓
+Simulated device policy package
+```
+
+You will run every service on the Arm cloud instance. The dashboard uses a simulated device, so this workflow doesn't connect to or move a physical robot.
+
+## Locate the exported actor
+
+Set `ACTOR_OUTPUT` to the artifact printed by `export_mappo_actor.py` in the training Learning Path. Replace the example filename with the frame count and source-checksum suffix from your export:
+
+```bash
+export ACTOR_OUTPUT="$HOME/mappo_actor_exports/mappo_actor_3agent_1910000_a1b2c3d4e5f6.npz"
+test -f "$ACTOR_OUTPUT" && echo "Actor found: $ACTOR_OUTPUT"
+```
+
+The command prints the complete actor path. If it prints nothing, correct `ACTOR_OUTPUT` before continuing.
+
+## Clone the MAPPO demo
+
+Clone the repository that contains the Device Connect dashboard:
+
+```bash
+export MAPPO_DEMO="$HOME/mappo-arm-cloud-physical-ai"
+git clone --filter=blob:none --no-checkout --depth 1 \
+    https://github.com/armwaheed/mappo-arm-cloud-physical-ai.git \
+    "$MAPPO_DEMO"
+git -C "$MAPPO_DEMO" fetch --depth 1 origin \
+    40d9be795da4e06725a1fc515ed5d3a6a9e7e5c1
+git -C "$MAPPO_DEMO" checkout --detach \
+    40d9be795da4e06725a1fc515ed5d3a6a9e7e5c1
+```
+
+The commit is pinned so that the commands and dashboard interface remain consistent with this Learning Path.
+
+## Install the dashboard dependencies
+
+Create a separate Python environment for the dashboard. Device Connect needs Python 3.11 or later; the Arm cloud environment from the training Learning Path provides Python 3.12:
+
+```bash
+python3.12 -m venv "$HOME/venvs/mappo-dashboard"
+source "$HOME/venvs/mappo-dashboard/bin/activate"
+python -m pip install --upgrade pip
+python -m pip install \
+    "device-connect-edge==0.2.5" \
+    "device-connect-agent-tools==0.2.5" \
+    "aiohttp==3.14.3" \
+    "numpy==2.4.6" \
+    "Pillow==12.3.0"
+```
+
+Verify the architecture, Python version, and Device Connect package versions:
+
+```bash
+python - <<'PY'
+import platform
+import sys
+from importlib.metadata import version
+
+print("Architecture:", platform.machine())
+print("Python:", sys.version.split()[0])
+print("Device Connect edge:", version("device-connect-edge"))
+print("Device Connect agent tools:", version("device-connect-agent-tools"))
+PY
+```
+
+The output is similar to:
+
+```output
+Architecture: aarch64
+Python: 3.12.x
+Device Connect edge: 0.2.5
+Device Connect agent tools: 0.2.5
+```
+
+## Check the actor interface
+
+The dashboard accepts the actor only if it contains the expected arrays and metadata. Run the same inspection that the dashboard applies after a model download:
+
+```bash
+cd "$MAPPO_DEMO/dashboard"
+python - <<'PY'
+import json
+import os
+
+from model_store import inspect_model
+
+report = inspect_model(os.environ["ACTOR_OUTPUT"])
+print(json.dumps(report.as_dict(), indent=2))
+raise SystemExit(0 if report.loadable else 1)
+PY
+```
+
+A compatible actor reports values similar to:
+
+```output
+{
+  "name": "mappo_actor_3agent_1910000_a1b2c3d4e5f6.npz",
+  "loadable": true,
+  "problems": [],
+  "trained_lidar_range_vmas": 0.35,
+  "rays": 12,
+  "training_frames": 1910000,
+  "training_n_agents": 3
+}
+```
+
+The command exits with a nonzero status if the actor can't be loaded. Don't continue if `problems` contains an unexpected array shape or LiDAR feature count.
+
+## Create a disposable policy package
+
+Arming a model changes `config.json` in the policy package. Copy the package to a temporary directory so the repository remains unchanged:
+
+```bash
+DASHBOARD_PACKAGE="$(mktemp -d -t mappo-dashboard-policy.XXXXXX)"
+export DASHBOARD_PACKAGE
+cp -a "$MAPPO_DEMO/policy/." "$DASHBOARD_PACKAGE/"
+echo "Dashboard package: $DASHBOARD_PACKAGE"
+```
+
+## What you've accomplished
+
+You have installed the dashboard dependencies, verified the exported actor contract, and created a disposable policy package. Next, you will start the Device Connect services and open the dashboard.

content/learning-paths/servers-and-cloud-computing/use-mappo-device-connect-dashboard/2-launch-dashboard.md

@@ -0,0 +1,78 @@
+---
+title: Launch the Device Connect dashboard
+description: Start the MAPPO model server, simulated device driver, and browser dashboard on the Arm cloud instance.
+weight: 3
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Forward the dashboard port
+
+The dashboard has no login. Keep it bound to the cloud instance's loopback interface and use SSH port forwarding instead of exposing it to the internet.
+
+Open another terminal on your local computer and connect to the cloud instance. Replace the username and address with your SSH details:
+
+```bash
+ssh -N -o ExitOnForwardFailure=yes \
+    -L 8080:127.0.0.1:8080 \
+    ubuntu@<cloud-instance-address>
+```
+
+Keep this SSH connection open while you use the dashboard.
+
+## Start the dashboard services
+
+Return to the SSH terminal where you set `ACTOR_OUTPUT`, `MAPPO_DEMO`, and `DASHBOARD_PACKAGE`. Start the dashboard without `--allow-motion`:
+
+```bash
+cd "$MAPPO_DEMO"
+./dashboard/start-dashboard.sh \
+    --python "$HOME/venvs/mappo-dashboard/bin/python" \
+    --package "$DASHBOARD_PACKAGE" \
+    --models-dir "$(dirname "$ACTOR_OUTPUT")"
+```
+
+The launcher starts three processes:
+
+- The model server publishes the `.npz` files in the actor export directory
+- The simulated driver exposes model-management functions through Device Connect
+- The web server presents the fleet and checkpoint controls in your browser
+
+The output ends with lines similar to:
+
+```output
+http://127.0.0.1:8080
+fleet       sim
+motion      DISABLED (status and checkpoints only).
+Ctrl-C stops all three.
+```
+
+{{% notice Note %}}
+The launcher deliberately leaves motion disabled. The simulated device is sufficient to validate the model distribution and selection workflow.
+{{% /notice %}}
+
+## Open the dashboard
+
+On your local computer, open [the forwarded Device Connect dashboard](http://127.0.0.1:8080/) in a browser.
+
+The following screenshot shows the dashboard's complete multi-robot layout. Use it to locate the **Fleet**, **Checkpoints on the robot**, and **Load from Cloud AI** panels. It was captured from a different, motion-enabled demonstration, so its header states **MOTION ENABLED** and **MESH DOWN**.
+
+![Arm Device Connect dashboard showing the robot fleet, motion controls, camera feed, installed MAPPO checkpoints, and Cloud AI model source. Use the Fleet and checkpoint panels as interface landmarks; this screenshot comes from a different demonstration with motion enabled and the mesh disconnected.#center](images/device-connect-dashboard.webp "Arm Device Connect dashboard interface reference")
+
+{{% notice Warning %}}
+Don't reproduce the motion state shown in the screenshot. Your simulation-only session must show **MESH UP** and **MOTION DISABLED** before you continue.
+{{% /notice %}}
+
+Confirm that the interface shows:
+
+- **MESH UP** in the header
+- `mappo-sim` with a **LIVE** state in the **Fleet** table
+- **MOTION DISABLED** in the header
+- `mappo-sim` selected under **Focus**
+
+The fleet row proves that the browser server discovered the simulated driver through the Device Connect mesh. The header also confirms that this run cannot issue motion commands.
+
+## What you've accomplished
+
+You have started the model server, Device Connect driver, and dashboard without exposing an unauthenticated port or enabling motion. Next, you will load and arm your exported actor through the dashboard.

content/learning-paths/servers-and-cloud-computing/use-mappo-device-connect-dashboard/3-load-actor.md

@@ -0,0 +1,84 @@
+---
+title: Load and validate the MAPPO actor
+description: Use the dashboard to download the exported actor, arm it for the simulated device, and run an inference smoke test.
+weight: 4
+
+### FIXED, DO NOT MODIFY
+layout: learningpathall
+---
+
+## Browse the model source
+
+The launcher advertises the local model server to the simulated device. The browser asks the device to browse that source, so the request follows the same Device Connect path used by a remote deployment.
+
+In the **Load from Cloud AI** panel, confirm that **Source** shows **local checkpoint server — local model server**. Select **Browse** if the actor list hasn't appeared automatically.
+
+The actor list shows your `.npz` filename and the message **served by mappo-model-server**. This response confirms that the simulated device can reach the model source.
+
+## Load the actor onto the simulated device
+
+Select **Use** beside your exported actor. Its address appears in the first **Source** field.
+
+Enter `trained_mappo_actor_part2.npz` in **Install as**. The new name prevents a collision if the disposable package already contains a checkpoint with the exporter's default filename.
+
+Select **Load onto robot**. In this simulation-only workflow, the destination is the disposable policy package rather than physical hardware.
+
+The result reports these checks:
+
+```output
+loaded trained_mappo_actor_part2.npz
+sha256         <actor checksum>
+rays           12
+trained range  0.35
+runnable now   yes
+
+Not armed. Arm it in the table above when you want the next run to use it.
+```
+
+The filename and checksum depend on your actor. Don't continue unless the result says `runnable now   yes`.
+
+## Arm the actor
+
+Find the downloaded actor in **Checkpoints on the robot**. It should have the **ready** state.
+
+Select **Arm** beside the actor. The state changes to **armed**, and the **Armed checkpoint** column in the fleet row shows the same filename.
+
+Loading and arming are separate operations. The dashboard inspects a downloaded file before it changes `model_path`, and an armed model takes effect only when the next policy process starts.
+
+Press the **E** key to open the event drawer. Look for the `model downloaded` and `checkpoint armed` events. These events record both changes made through Device Connect.
+
+## Stop the dashboard
+
+Return to the cloud SSH terminal running `start-dashboard.sh` and press **Ctrl+C**. The launcher stops the model server, simulated driver, and web server together.
+
+The output names each process as it stops:
+
+```output
+stopping checkpoint server
+stopping driver
+stopping dashboard
+```
+
+## Run an inference smoke test
+
+The disposable policy package now points to the actor you armed. Run its installation check:
+
+```bash
+source "$HOME/venvs/mappo-dashboard/bin/activate"
+python "$DASHBOARD_PACKAGE/basic_test.py"
+```
+
+The output identifies your checkpoint and ends with:
+
+```output
+checkpoint       trained_mappo_actor_part2.npz
+trained on       1910000 frames, 3 agents
+ActionOutput(...)
+PASS
+```
+
+The action values depend on your actor. `PASS` confirms that the policy package loaded the armed arrays, constructed an 18-value observation, and completed one inference step.
+
+## What you've accomplished
+
+You have served an exported MAPPO actor, transferred it through Device Connect, armed it in a disposable policy package, and validated inference. The complete workflow used a simulated device and did not connect to or move physical hardware.

content/learning-paths/servers-and-cloud-computing/use-mappo-device-connect-dashboard/_index.md

@@ -0,0 +1,64 @@
+---
+title: Load a MAPPO policy with the Arm Device Connect dashboard
+description: Use the Arm Device Connect dashboard to distribute, select, and validate an exported MAPPO policy on an Arm cloud instance.
+minutes_to_complete: 45
+
+who_is_this_for: This Learning Path is for machine learning developers who have exported a MAPPO actor and want to validate its deployment workflow through a browser-based Device Connect dashboard.
+
+draft: true
+cascade:
+    draft: true
+
+learning_objectives:
+    - Set up the MAPPO Device Connect dashboard on an Arm cloud instance.
+    - Serve an exported MAPPO actor to a simulated device through Device Connect.
+    - Load, arm, and validate the selected actor without connecting physical hardware.
+
+prerequisites:
+    - An Arm-based Ubuntu 24.04 cloud instance with SSH access, `sudo` privileges, and internet access.
+    - The actor-only `.npz` artifact created in the [MAPPO training Learning Path](/learning-paths/servers-and-cloud-computing/train-mappo-navigation-arm-cloud/).
+    - A local browser and permission to forward port 8080 through SSH.
+
+author: Waheed Brown
+
+generate_summary_faq: true
+rerun_summary: false
+rerun_faqs: false
+
+### Tags
+skilllevels: Advanced
+subjects: ML
+armips:
+    - Neoverse
+tools_software_languages:
+    - Arm Device Connect
+    - MAPPO
+    - Python
+    - NumPy
+operatingsystems:
+    - Linux
+
+further_reading:
+    - resource:
+        title: Train a MAPPO navigation policy on Arm cloud
+        link: /learning-paths/servers-and-cloud-computing/train-mappo-navigation-arm-cloud/
+        type: website
+    - resource:
+        title: MAPPO Arm cloud Physical AI demo
+        link: https://github.com/armwaheed/mappo-arm-cloud-physical-ai
+        type: website
+    - resource:
+        title: Arm Device Connect repository
+        link: https://github.com/arm/device-connect
+        type: website
+    - resource:
+        title: BenchMARL repository
+        link: https://github.com/facebookresearch/BenchMARL
+        type: website
+
+### FIXED, DO NOT MODIFY
+# ================================================================================
+weight: 1
+layout: "learningpathall"
+learning_path_main_page: "yes"
+---

content/learning-paths/servers-and-cloud-computing/use-mappo-device-connect-dashboard/_next-steps.md

@@ -0,0 +1,8 @@
+---
+# ================================================================================
+#       FIXED, DO NOT MODIFY THIS FILE
+# ================================================================================
+weight: 21                  # Set to always be larger than the content in this path to be at the end of the navigation.
+title: "Next Steps"         # Always the same, html page title.
+layout: "learningpathall"   # All files under learning paths have this same wrapper for Hugo processing.
+---