Anomaly Detection (Beta)

The Transitive Robotics anomaly detection (AD) capability watches your robot's topics and tells you when something is wrong. No threshold tuning, no hardcoded rules. It learns what normal looks like on your robot, then flags anomalies in real time.

Currently ROS 1 and ROS 2 (all distributions) are supported. Reach out if you are using something else and are interested in using this capability. Contact details are in the Support section below.

1. Discover: The capability automatically discovers the topics coming from the robot, classifies them and makes suggestions on which ones would be good for an AD model.
2. Define: Add topics you want to track in the AD model.
3. Collect: After you have defined your AD model, the capability collects data from the selected topics and uploads it to the server to be used for training.
4. Train: An AD model is then trained from the data that was collected and deployed back out to your robot.
5. Detect: The capability then uses the AD model to watch the robot's topics and will alert you to problems as they arise.

Anomaly Detection is in beta and under active development. It's improving fast and your participation genuinely shapes where it goes. All feedback and questions are welcome.

So please reach out; whether you want a hand getting set up or are interested in working directly with Target Node on your fleet. We'd love to hear what you're building.

• Email: anomaly-beta@targetnode.ai