Flywheel is a cloud‑based remote teleoperation platform that lets robotics engineers control off‑the‑shelf and custom robots via an intuitive web UI. It automatically logs operator actions and sensor data, generating rich datasets that can be exported to train and improve autonomous algorithms, while supporting standard middleware like ROS and customizable data publishers.
Funding
Funding not disclosed
Founders
Product
Problem
Robotics developers often spend extensive time manually operating and testing robots on-site, and they lack efficient tools for collecting real-world interaction data needed to improve autonomous capabilities. This slows product development, increases costs, and limits the amount of high-quality data available for training autonomous systems.
Solution
Flywheel provides a cloud‑based remote teleoperation platform that lets users control both off‑the‑shelf and custom robots through an intuitive web interface. Operators can issue commands, monitor robot status, and define custom events or tags, while the system automatically logs all interactions in the background. The collected logs form a rich dataset that can be used to train and refine autonomous algorithms, accelerating the transition from manual control to autonomy. Flywheel’s modular architecture supports customizable publishers, enabling integration with existing robot middleware and data pipelines. By centralizing control and data capture, the platform reduces the time and effort required to bring robots to market.
Target Audience
Flywheel is aimed at robotics engineers, R&D teams, and product developers who need a fast, low‑overhead way to operate robots remotely and gather high‑quality data for autonomy development.
Features
- Web UI with drag‑and‑drop controls for real‑time teleoperation of diverse robot platforms
- Automatic event tagging and continuous background logging of sensor streams and operator actions
- Configurable publishers that forward telemetry to user‑specified endpoints or cloud services
- Data export tools for building large‑scale datasets to train machine‑learning models
- Compatibility with standard robot middleware (e.g., ROS) and custom hardware integrations
- Scalable cloud infrastructure that supports multiple concurrent operators and robots