swarmit 0.8.0rc1

Run Your Own Robot Swarm Testbed.

SwarmIT

SwarmIT provides a embedded C port for nRF53 as well as Python based services to easily build and deploy a robotic swarm infrastructure testbed. ARM TrustZone is used to create a sandboxed user environment on each device under test, without requiring a control co-processor attached to it.

Features

• Experiment management: start, stop, monitor and status check
• Deploy a custom firmware on all or on a subset of robots of a swarm testbed
• Resilient robot state: even when crashed by buggy user code, the robot can be reprogrammed remotely and wirelessly

Usage

Get the code

Swarmit depends on the DotBot-libs and Mari repositories. They are included in the codebase as Git submodules.

Use the following command to clone the Swarmit codebase locally:

git clone --recurse-submodules https://github.com/DotBots/swarmit.git

Embedded C code

SwarmIT embedded C code can be built using Segger Embedded Studio (SES). Use Tools > Package manager to install the CMSIS 5 CMSIS-CORE, CMSIS-DSP and nRF packages.

To provision a device, follow the following steps:

1. open netcore.emProject and bootloader.emProject (or bootloader.emProject depending on your robot version) in SES
2. build and load the netcore application on the nRF53 network core,
3. build and load the bootloader application on the nRF53 application core.

The device is now ready.

Lighthouse calibration

The bootloader compiles in a Lighthouse v2 homography from device/bootloader/Source/lh2_calibration.h. That file is auto-generated by the dotbot-lh2-calibration tooling — see its README for the calibration procedure. Quick recap:

# 1. Calibrate against the four reference points (-d is the square side in mm)
dotbot-calibration -p <serial-port> -d 500   # 50 cm square, ~2.5 m × 2.5 m arena

# 2. Export the calibration as a C header next to the bootloader sources
dotbot-calibration-exporter device/bootloader/Source

Rebuild and reflash the bootloader for the new calibration to take effect.

For already-flashed robots there's no need to rebuild — push the new calibration over the air with swarmit calibrate-lh2, see Pushing an LH2 calibration over the air below.

Gateway

The communication between the computer and the swarm devices is performed via a gateway board connected via USB to the computer.

This gateway uses the mari network stack and must run the Mari gateway firmware.

The documentation to setup a Mari gateway is located here.

Python CLI script

The Python CLI script provides commands for flashing, starting and stopping user code on the device, as well as monitoring and checking the status of devices in the swarm.

The Python CLI script connects via a virtual COM port to the gateway connected to the computer.

The Python CLI script is available on PyPI. Install it using:

pip install swarmit

Print usage using swarmit --help:

Usage: swarmit [OPTIONS] COMMAND [ARGS]...

Options:
  -c, --config-path FILE      Path to a .toml configuration file.
  -p, --port TEXT             Serial port to use to send the bitstream to the
                              gateway. Default: /dev/ttyACM0.
  -b, --baudrate INTEGER      Serial port baudrate. Default: 1000000.
  -H, --mqtt-host TEXT        MQTT host. Default: localhost.
  -P, --mqtt-port INTEGER     MQTT port. Default: 1883.
  -T, --mqtt-use_tls          Use TLS with MQTT.
  -n, --network-id TEXT       Marilib network ID to use. Default: 0x1200
  -a, --adapter [edge|cloud]  Choose the adapter to communicate with the
                              gateway. Default: edge
  -d, --devices TEXT          Subset list of device addresses to interact with,
                              separated with ,
  -v, --verbose               Enable verbose mode.
  -V, --version               Show the version and exit.
  -h, --help                  Show this message and exit.

Commands:
  calibrate-lh2  Send LH2 calibration data to the robots.
  flash          Flash a firmware to the robots.
  message        Send a custom text message to the robots.
  monitor        Monitor running applications.
  reset          Reset robots locations.
  start          Start the user application.
  status         Print current status of the robots.
  stop           Stop the user application.

Pushing an LH2 calibration over the air

Once a robot is flashed and connected to the Mari network, you can update its Lighthouse v2 homography without re-flashing the bootloader. The CLI pushes a calibration file produced by dotbot-lh2-calibration (default location ~/.dotbot/calibration.out) over Mari; the network core writes it to the config page in flash and the SoC resets so the bootloader loads the new homographies on the next boot.

# Push the most recent calibration to every ready device on network 0xA000
swarmit -n 0xA000 calibrate-lh2 ~/.dotbot/calibration.out

# Or target a subset (-d takes a comma-separated list of device addresses)
swarmit -n 0xA000 -d BC3D3C8A2A6F8E68 calibrate-lh2 ~/.dotbot/calibration.out

swarmit serve

swarmit serve is the unified FastAPI backend. Two deployment presets, same binary:

• Local-dev convenience — swarmit serve --local. Binds 127.0.0.1, no JWT auth, no records DB. The local CLI auto-discovers it at localhost:8001 and routes commands through HTTP/SSE (sub-50 ms cold-start) instead of building a fresh in-process Controller per invocation. Pass --no-server to force the legacy in-process path.

• Shared service — swarmit serve (default). Binds 0.0.0.0, JWT required, JWT records DB on. Used on a testbed server reachable by operators and remote CLIs. React UI mounted on the same port.

pip install swarmit[dashboard]                # includes the serve subcommand
swarmit -n 0x1234 serve --local &             # local-dev preset

# Same CLI, now answered by the server:
swarmit status                                # live status, sub-50 ms
swarmit flash sample.bin                      # streaming OTA progress (SSE)
swarmit status -w                             # SSE-driven Rich Live table
swarmit monitor                               # streams SWARMIT_EVENT_LOG

# Override the server endpoint:
SWARMIT_SERVER_URL=http://127.0.0.1:9001 swarmit status

swarmit serve --local refuses to bind to any address other than localhost — using --bind-host 0.0.0.0 with --local is rejected. Cross-machine deployment requires the default JWT-enabled mode.

swarmit-server is kept as a deprecated standalone console_script that behaves identically to swarmit serve. New scripts should use swarmit serve.

python -m swarmit.dashboard.main is kept as a deprecated alias that forwards to swarmit-server.

Control Tower Dashboard

The Control Tower is a web-based platform (backend and frontend) that enables users to manage and monitor the testbed remotely. It provides an interface for reserving timeslots, inspecting the live status of all DotBots, and supervising experiments. The platform displays each device’s position and operational state, and offers mechanisms to flash firmware, start or stop experiments, and oversee ongoing activity across the testbed.

Setup

1. Download all requirements.

pip install swarmit[dashboard]

2. Generate a private and public key for the JWT

# Create the data directory
mkdir -p .data

# Generate Ed25519 private key
openssl genpkey -algorithm Ed25519 -out .data/private.pem

# Extract the public key
openssl pkey -in .data/private.pem -pubout -out .data/public.pem

Running the Dashboard

After the initial setup (required only once), you can launch the dashboard. With the edge adapter (default — gateway over USB serial):

python3 -m swarmit.dashboard.main --http-port 8080 --open-browser

With the cloud adapter (gateway reached over an MQTT broker):

python3 -m swarmit.dashboard.main -a cloud -n 0x1234 \
    -H broker.example.com -P 8883 -T \
    --http-port 8080 --open-browser

Or use a TOML config file (recommended once you have one):

python3 -m swarmit.dashboard.main -c swarmit-argus.toml -n 1234 \
    --http-port 8080 --open-browser

Note: the dashboard opens its own gateway connection. If swarmit serve is already running and owns the gateway (serial port, especially), stop it before starting another instance.

Access the dashboard at https://localhost:8080

NOTE: Your dashboard CLI options may differ depending on your example or environment.

The map view draws a graph-paper grid (minor cells every d mm, major squares every 5d mm — one Lighthouse v2 coverage area) and marks the four LH2 reference points used by dotbot-calibration. For a single-LH arena the grid auto-infers d from --map-size (min(width, height) / 5). For multi-LH arenas where the arena extends past LH0's coverage, pass it explicitly:

# Two stacked LHs (d=200 mm), arena 1000x1800 mm with 200 mm overlap
python3 -m swarmit.dashboard.main \
    --map-size 1000x1800 --calibration-distance 200