This document discusses using Kubernetes to manage a distributed robotics system across edge and cloud computing environments. It introduces Tomoya Fujita from Sony Corporation and provides an overview of robot operating systems, Sony's robotics architecture, and the problems of developing applications across different edge and cloud environments. It then proposes using Kubernetes to provide a common base architecture and development environment for robotics applications across edge nodes, cloud nodes, and different hardware platforms. Key features discussed include running applications as containers, virtual device plugins, node discovery, and a circulatory functioning system.

1. R&D Center Tokyo Lab 29, Sec 2
Copyright 2020 Sony Corporation
Kubernetes
Robotics Distributed System
Tomoya Fujita
R&D Center
Sony Corporation

2. Agenda
 Self-Introduction
 Robotics Operating System
 Sony Robotics Architecture
 Problem / Pain
 Kubernetes Edge Robotics Cluster System

3. Who am I?
 Tomoya Fujita
• https://www.linkedin.com/in/tomoya-fujita-5bb656b6/
• https://github.com/fujitatomoya
 2016 – Sony Corporation
– System Software Engineer / Architect
– Robotics Product Development
– Robotics System Engineer
– Cloud Framework (Kubernetes)
 Robotics Relate Work
– https://vimeo.com/293292255 (ROSCon2018)
– https://vimeo.com/378682623 (ROSCon2019)
– https://www.theconstructsim.com/openai-ros-workshop-roscon/rosdevcon19-
tomoya-fujita/ (ROSDevcon2019 Hands-On)

4. Robotics Operating System
 Robotics Eco-System
 Open Source
 Meta-Operating System
 Publisher/Subscriber Architecture
 Simulation / Visualization
 Debug Utility
 Hardware Abstraction

5. Hardware
R e c o g n i t i o n
B e h a v i o r
C o n t r o l
S i t u a t i o n a l A w a r e n e s s
B e h a v i o r
P l a n n i n g
Int elligent Processing
（Edge）
User/Environment
Int elligent
Processing
（Cloud）
Sensing Mechatronics
IN
OUT
Edge Computing System

6. IoT
Devices
Robot
App
Robot
App
Cloud
IoT
Devices
Robots
App App
App App
App
App
Common Base Architecture for Everyone
What’s the problem?
• Different Architecture for Cloud and Edge device.
• Take time to setup environment and run application.
• IoT Device is really statistic implementation.
What we wanna do?
• Common Base Architecture for everyone.
• During Development, we can test every service on
single device.
• More Flexibility for Application. More like EcoSystem.
Cloud
Robots
Cloud
App
Cloud
App
Sensor
App
Sensor
App

7. Edge Node
Application
Container
System
Ext APIs
Application
System
Services
Kernel & Drivers
Output
Edge Node
Kernel & Drivers
Edge Cluster Network
(Slave Cluster)
EdgeCluster
Master
Capabilities
NodeController
NodeController
Application
Container
System
Ext APIs
Application
System
Services
Output
Capabilities
Cloud Cluster Network
(Host Cluster)
Federation
Cloud Node
Application
Container
Application
NodeController
Kernel & Drivers
GPUAccess
Cloud Node
Application
Container
Application
NodeController
Kernel & Drivers
GPUAccess
CloudCluster
Master
x86 arm64
Accelerator
Accelerator

8. LAN
Kubernetes Master (x86) Kubernetes Slave (arm64) Kubernetes Slave (arm64)
CNI – Weave (Layer 2 Emulation)
KubernetesAPIServer
Kubelet
Kubelet
Kubelet
ApplicationPod
Dashboard
Listener
ApplicationPod
Talker
1Hz
ApplicationPod
Listener

9. What else?
 Virtual Device Plugin.
 Node Discovery.
 Circulatory Functioning System.
 Contribution.
 https://github.com/kubernetes/kubernetes/issues/86539