Introducing Amazon SageMaker Kubeflow Reinforcement Learning Pipelines for Robotics
summary
Orchestrates reinforcement learning from perception to controls using SageMaker + Kubeflow pipelines and AnyScale’s Ray RLlib to scale trainers and simulators, persist checkpoints to S3, and push logs to CloudWatch. Two customer wins: Woodside Energy trains a dual‑arm navigator Ripley in AWS RoboMaker and shows one‑to‑one sim‑to‑real behavior; GE Aviation trains a Fanuc manipulator in PyBullet then deploys to the factory floor. Benchmarked Ray on Pong hitting 8M steps in ~23–30 minutes depending on cluster shape.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:03Outline
- 2:05Reinforcement Learning Workflow
- 8:00Reinforcement Learning
- 10:53Kubeflow Pipelines
- 17:48Customer Use Cases
- 19:25GE Use Case
related talks
31mUtilizing Reinforcement Learning Techniques for Supply Chain with Amazon SageMaker
34mBuilding End-to-End Machine Learning Workflows with Kubernetes...
24mGet Rolling on Machine Learning
29mGet rolling with ML on AWS DeepRacer
28mBuild smart robots and operate them easily with AWS Robotics
54mScaling Machine Learning Pipelines in Cloud
37mAWS DeepRacer: Train, Evaluate, and Tune
27mReplicate AWS DeepRacer architecture to conquer the track with SageMaker