Fleet Telemetry

This document describes the fleet monitoring architecture for production deployments of the 3WE Robot Platform.

Architecture

┌─────────────┐     ┌─────────────┐     ┌─────────────────┐     ┌──────────┐
│  Robot #1   │     │  Robot #2   │     │  MQTT Broker    │     │ Grafana  │
│             │     │             │     │  (Mosquitto)    │     │          │
│ diagnostics ├────►│ diagnostics ├────►│                 ├────►│ Dashboard│
│ mqtt_bridge │MQTT │ mqtt_bridge │MQTT │ fleet/+/diag    │     │          │
└─────────────┘     └─────────────┘     └────────┬────────┘     └──────────┘
                                                  │
                                                  ▼
                                         ┌─────────────────┐
                                         │ Telegraf/InfluxDB│
                                         │ (Time Series DB) │
                                         └─────────────────┘

Components

1. Diagnostics Node (robot_diagnostics)

Runs on each robot's companion computer. Subscribes to platform health topics and publishes a unified DiagnosticArray at 1 Hz.

ros2 run robot_diagnostics diagnostics_node --ros-args \
  -p robot_id:=robot-001 \
  -p publish_rate_hz:=1.0

2. MQTT Bridge Node (mqtt_bridge_node)

Forwards /diagnostics messages to a fleet MQTT broker.

ros2 run robot_diagnostics mqtt_bridge_node --ros-args \
  -p broker_url:=mqtt://fleet-broker.local:1883 \
  -p username:=robot \
  -p password:=secret \
  -p robot_id:=robot-001 \
  -p topic_prefix:=fleet/ \
  -p qos:=1

3. MQTT Broker (Mosquitto)

Central broker collecting telemetry from all robots.

# mosquitto.conf
listener 1883
allow_anonymous false
password_file /etc/mosquitto/passwd
persistence true
persistence_location /var/lib/mosquitto/

4. InfluxDB + Telegraf

Telegraf subscribes to MQTT topics and writes to InfluxDB.

# telegraf.conf
[[inputs.mqtt_consumer]]
  servers = ["tcp://localhost:1883"]
  topics = ["fleet/+/diagnostics"]
  data_format = "json"
  json_time_key = "timestamp"
  json_time_format = "unix"
  tag_keys = ["robot_id"]

[[outputs.influxdb_v2]]
  urls = ["http://localhost:8086"]
  token = "$INFLUX_TOKEN"
  organization = "robot-platform"
  bucket = "fleet_telemetry"

MQTT Topic Schema

fleet/{robot_id}/diagnostics

Payload (JSON):

{
  "timestamp": 1705312200.123,
  "robot_id": "robot-001",
  "status": [
    {
      "name": "robot-001/system",
      "level": 0,
      "message": "Running",
      "values": {
        "uptime_s": "3621.5",
        "robot_id": "robot-001"
      }
    },
    {
      "name": "robot-001/battery",
      "level": 0,
      "message": "OK: 78%",
      "values": {
        "percentage": "78.2",
        "voltage_v": "7.84"
      }
    },
    {
      "name": "robot-001/safety",
      "level": 0,
      "message": "Normal operation",
      "values": {
        "estop_active": "false"
      }
    }
  ]
}

Diagnostic Levels

Level	Name	Meaning
0	OK	Operating normally
1	WARN	Degraded (e.g., low battery, E-stop active)
2	ERROR	Critical issue (e.g., battery <10%)
3	STALE	No data received (sensor offline)

Alerting Rules

Configure in Grafana or Alertmanager:

Condition	Severity	Action
Battery < 10%	Critical	Notify ops, initiate safe shutdown
E-stop active > 5 min	Warning	Investigate stuck robot
Topic rate = 0 for > 30s	Warning	Communication failure
Uptime reset unexpected	Info	Robot rebooted (check core dump)
No telemetry > 2 min	Critical	Robot offline

Quick Start (Docker Compose)

The docker-compose.yml in the project root includes an optional MQTT broker:

docker compose --profile fleet up -d

This starts: - Mosquitto MQTT broker on port 1883 - The robot companion stack (rosbridge + diagnostics + web UI)

For a full monitoring stack, add Telegraf + InfluxDB + Grafana to your fleet infrastructure.

Data Retention

Recommended retention policy: - Raw telemetry (1 Hz): 7 days - Aggregated (1 min): 90 days - Aggregated (1 hour): 2 years

Configure in InfluxDB bucket retention rules.

Security

• MQTT connections should use TLS (mqtts://) in production
• Each robot gets unique MQTT credentials (username = robot_id)
• Broker ACLs restrict each robot to its own topic subtree
• Fleet dashboard access requires separate authentication