Robots That Keep Learning: Edge-Cloud Lifelong Learning with KubeEdge Ianvs

Robots meet conditions their training data never covered: new roads, lighting, weather. Lifelong learning lets them keep improving, but running it at the edge raises practical questions. Where does inference run, when does a model update, and how do you know the update helped?

This talk shows how KubeEdge Ianvs and Sedna handle this. Using a robot scenario on image data, I walk through the lifelong learning loop: edge inference, detecting unseen data, cloud retraining, and redeployment. I cover what broke when I restored these examples after Ianvs core changed its dataset and paradigm contracts, and how to benchmark a lifelong learner so results can be reproduced.

You will leave knowing how edge-cloud lifelong learning is structured, how to evaluate it, and how to start contributing to KubeEdge SIG AI.