Edge analytics ranges from simple to sophisticated: a threshold check on a sensor, a rolling statistic on a gateway, or a trained model running on an edge gateway that flags a bearing fault from vibration data. In each case the raw stream stays local and the uplink carries what matters, which reduces bandwidth, shortens reaction time and keeps sensitive data on-site.
The trade-off is between local decisions and the global view. A gateway sees its own machines, not the fleet, so patterns that only appear across sites are missed unless some data still flows to the cloud. Most designs send summaries plus a sample of raw data, and update the edge logic through the same channel used for firmware. Edge computing explained goes into the architecture; edge AI covers the machine learning end.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.