Kubernetes Robotics Edge Cluster System
summary
Kubernetes runs ROS-based robot fleets so multiple robots connect directly, deploy apps as containers, and the team demonstrates lifecycle, recovery and scaling claims up to 10,000-node scale. Networking uses Weave Net CNI to support ROS multicast discovery. Hardware and host APIs attach via the device-plugin with virtual devices, while per-endpoint secrets and configmaps provide dynamic security enclaves. Device-plugin resource release is still missing and a KEP is in progress, but security attachment and basic distributed demos work.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:32Agenda
- 1:07Self-Introduction
- 2:02Sony Purpose
- 2:35General Background
- 3:20Robotics Background
- 4:42ROS
- 6:23Problems
- 7:10Goal
- 7:58Application Friendly
- 8:52Advantage
- 9:40Common Architecture
- 11:01Distributed System
- 12:29Security Enclaves
- 17:19Device-Plugin Open Issue
- 18:15Cluster Reconfiguration
- 20:34Plan
related talks
25mBuilding a PaaS for Robotics with Kubernetes
21mBuilding a PaaS for Robotics with Kubernetes
14mReliability Engineering for Hybrid Robot-Cloud Systems
29mIntegrate ROS into EdgeX - Startover of Edge Computing System
43mPatterns for cloud-enabled robots
36mScaling robots -
30mRobot Operating System (ROS): current and future capabilities on embedded systems
48mBuilding Robotics Applications at Scale using Open Source from Zero to Hero