Project GoHusky : a RL motion based Husky playground in Quad UW

This a a project adapted from Puppo, The Corgi - Unity Blog With update and new features of

1. Modern Ways of Using ML-Agents Library Function Calls
2. UW-Themed Scene with Husky
3. Project Setup
4. Training Script Provided

Modern Ways of Using ML-Agents Library Function Calls

• Describe the modern ways of using ML-Agents library function calls in your project.
• Initialize()
• OnEpisodeBegin()
• CollectObservations(VectorSensor sensor)
• OnActionReceived(ActionBuffers actions)
• FixedUpdate()

UW-Themed Scene with Husky

• Showcase the UW-themed scene with the Husky character.
• Demo

Project Setup

Step 1: Get Unity

• Download Unity from Unity's official website.

Step 2: Create a New Project

• Open Unity Hub.
• Click "New" to create a project.
• Enter a project name and location.
• Choose project settings (e.g., 2D or 3D).
• Click "Create."

Step 3: Import Packages

• Open the Unity editor.

• In the Unity editor, navigate to "Window" > "Package Manager."

• install the following packages:

• packages:

  ML-Agent, from "git+https://github.com/Unity-Technologies/ml-agents.git?path=com.unity.ml-agents", ver 3.0.0

  Cinemachine, ver 2.9.7

  Recorder, ver 4.0.2

Step 4: Play the Game

• You are all set, Click the "Play" button in Unity to start the game.

Caution: Adjust the Start Button Position

• Depending on your system (mac, Linux, Windows), the "START" button may appear off-screen.
• In the Unity Scene Hierarchy, locate "Canvas/TitlePanel/StartButton."
• In the Inspector, adjust PosX and PosY properties as needed.

Training Script(adopted from Hugging Face)

behaviors:
  CorgiFetch:
    trainer_type: ppo
    hyperparameters:
      batch_size: 2048
      buffer_size: 20480
      learning_rate: 0.0003
      beta: 0.005
      epsilon: 0.2
      lambd: 0.95
      num_epoch: 3
      learning_rate_schedule: linear
    network_settings:
      normalize: true
      hidden_units: 512
      num_layers: 3
      vis_encode_type: simple
    reward_signals:
      extrinsic:
        gamma: 0.995
        strength: 1.0
    checkpoint_interval: 200000
    keep_checkpoints: 15
    max_steps: 2e7
    time_horizon: 1000
    summary_freq: 50000