Concept Pet

Train an AI puppy with rewards — and learn how reinforcement learning works.

Concept Pet is an iOS app that teaches AI and reinforcement learning concepts through hands-on gameplay. You train a virtual puppy by giving treats (positive rewards) or saying “no” (negative rewards), and watch it get smarter over time using a real Q-learning algorithm.

Features

Training Levels

Each level teaches a different reinforcement learning concept:

Level	Concept	Description
Ball Fetch	Reward Shaping	Guide your pup toward the ball with incremental rewards
Sit Command	Reinforcement	Teach your dog to sit on command through feedback
Maze Runner	Exploration vs. Exploitation	Help your pup decide when to try new paths or stick with what works
Patrol Route	Policy Learning	Train a complete patrol plan across multiple waypoints
Stinky Sock	Avoidance Learning	Chase the ball while learning to avoid danger

Brain Map

Visualize your puppy’s Q-table in real time. Each cell shows which direction the pup would move and how confident it is — watch the brain map evolve as your pup learns from your feedback.

Customize Your Pup

Upload a photo of your real dog and use Apple’s Image Playground to generate a unique AI-styled avatar for your in-app pet. (Requires iOS 18.1+ with Apple Intelligence enabled.)

Enthusiast Mode

Toggle advanced settings to fine-tune treat and penalty reward values per level for deeper experimentation with the RL algorithm.

Tech Stack

• SwiftUI — UI framework
• SpriteKit — Game rendering (grid environments, pet animation)
• Q-Learning — Tabular reinforcement learning with configurable learning rate, discount factor, and epsilon-greedy exploration
• Swift 6 — Full concurrency safety
• Image Playground — Apple Intelligence integration for pet customization

Requirements

• iOS 17.0+
• iPhone or iPad
• Xcode 16+ to build

Project Structure

Concept Pet.swiftpm/
├── MyApp.swift              # App entry point
├── ContentView.swift        # Root navigation
├── HomeView.swift           # Home screen with pet avatar
├── LevelSelectView.swift    # Level picker
├── LevelIntroView.swift     # Pre-level concept explainer
├── GameView.swift           # Training UI (treat/no buttons)
├── GameScene.swift          # SpriteKit game scene
├── GameState.swift          # Central state + level definitions
├── BrainMapView.swift       # Q-table visualization
├── RLAgent.swift            # RL agent (epsilon-greedy + Q-learning)
├── QTable.swift             # Q-table implementation
├── RLEnvironment.swift      # Environment protocol
├── FetchEnvironment.swift   # Ball fetch grid environment
├── SitEnvironment.swift     # Sit command environment
├── MazeEnvironment.swift    # Maze navigation environment
├── PatrolEnvironment.swift  # Waypoint patrol environment
├── SockEnvironment.swift    # Fetch + avoidance environment
├── PetNode.swift            # SpriteKit pet node
├── SpriteLoader.swift       # Sprite sheet loading
├── AppSettings.swift        # User preferences + reward tuning
├── SettingsView.swift       # Settings UI
├── Theme.swift              # Colors and typography
└── Resources/
    └── sprite123.png        # Pet sprite sheet

Building

1. Open Concept Pet.swiftpm in Xcode 16+
2. Select an iPhone or iPad simulator
3. Build and run (Cmd + R)

Privacy Policy

Privacy Policy

License

All rights reserved.