Anomaly Detection (Alpha)

Faros — ROS Anomaly Detection (Alpha)

Faros watches the topics your robot publishes and flags anomalies in real time. It trains a model for on a group of related signals off the robot, then runs those models on the robot to score incoming data.

This is a Transitive capability. Installing it from the marketplace puts the Faros agent on a device and starts sending anomaly data to the Faros server.

• Runs inference at the edge. The model scores a sliding window of recent samples and flags both gradual drift and sudden spikes. Inference generally runs at 5 Hz.
• Discovers topics. It lists every ROS topic and field on the robot so you can choose which ones to monitor.
• Collects data. It records data and uploads it to your Faros server for training.
• Provides an operator dashboard. The capability ships a React dashboard that you embed in whatever tool you use to manage robots.
• Sends a heartbeat. Every 30 seconds the agent reports its health. Operators can read this without connecting to the device.

ALL

If a device's credentials are ever wrong or missing or you cannot view your agent in the browser, simply restart the capability on that device to fix it.

This is an alpha release. Email shane@shanehill.com for installation help or to report issues.