kubecon-ai-demos

Rock Paper Scissor Game

Goals

• Build an ML inference library to interpret human hand gestures

• Build a computer UI for gameplay

  • 3..2..1..go
  • Visual display of computer move
  • Reset game
  • Humans play multiple rounds (i.e. best of 10)

• Build a humanoid robotic arm to show computer moves

• All gameplay components run in containers

  • camera feed (rtsp feed preferred to onboard integration)
  • Inference algorithm
  • Robot arm gestures
  • Gameplay UI
  • Message bus

• Deliver the solution as an open source project in a public git repo

• Demo the cluster(s) connected via Rancher

• Demo gitops updates/rollbacks for various components

System

• Jetson Xavier NX Dev Kit
• Raspberry Pi camera (alternative streaming webcam)
• Robotic Arm (https://github.com/mak3r/edge-robot-demo)
• Rancher Management Server (running where?)

Stretch Goals

• Build a secondary library to do predictive gameplay for the computer player
• Receive camera feed from webcam interfaces remotely
• Akri integration for camera, robot, jetson

Demo

The “demo” is broken into two basic parts. Part 1 is an interactive game of Rock Paper Scissors in which human players attempt to win against the computer. Part 2 will be a presentation that includes both gameplay and the unpacking of NUCs prepared for the OT environment and dropped in to demonstrate how AI/ML, Edge systems and Secure Device Onboarding can be integrated with cloud native technologies to provide complete, secure and remotely manageable end-to-end solutions.

The interactive demo will have 5-6 physical components on the desktop. These include:

• 3 NUCs as the Kubernetes cluster
• A web cam for input of human player gestures
• A robotic arm for expression of the computer gestures
• A device to start/stop/reset the game