Simple Q - ML Framework
Simple Q is a free, modular framework for building and training AI agents in Unity using the Q Learning algorithm, without requiring additional assets or Python knowledge.
Simple Q is a free, modular framework for building and training AI agents in Unity using the Q Learning algorithm, without requiring additional assets or Python knowledge.* This page contains affiliate links, meaning we may earn a small commission if you purchase something through them, at no extra cost. FREE
- Category:
- Tools › Ai-ml-integration
- Developer:
- CircuitZ
- Price:
- FREE
- Favorites:
- 14
- Supported Unity Versions:
- 2022.3.14 or higher
- Current Version:
- 1.0.0
- Download Size:
- 1.24 MB
- Last Update:
- May 29, 2024
- Description:
- Simple Q is a comprehensive AI/ML framework that allows developers to build, train, and deploy working AI agents in Unity without needing to download additional assets or learn Python. The framework includes a range of features such as Q Learning Algorithm + QTable, Replay Buffer Experience, Exploration Vs Exploitation, Dynamic Decay, Shared Data Component, Persistent Data, Prioritized Learning, Buffer Replay Removal Policy, Attribute Pairs, and Little Q.
The framework is fully documented, with all code accompanied by comments explaining their functionality. Three examples are included to get started. With Simple Q, developers can not only build and train AI agents but also learn the basics of AI and Machine Learning, allowing them to create amazing things.
The framework includes:
* Q Learning Algorithm + QTable
* Replay Buffer Experience
* Exploration Vs Exploitation (two methods available)
* Dynamic Decay (two common methods, and a new example method)
* Shared Data Component (or Individual Experience without)
* Persistent Data
* Prioritized Learning (with three working examples to choose from)
* Buffer Replay Removal Policy (with four working examples to choose from)
* Attribute Pairs (For better Experience Sampling)
* Little Q (a very basic setup for Q Learning - great for learning the basics)
* Three agents/environments (Virtual Battle Bot, 2D Navigator, Dodge Bot (3D))
Simple Q's reinforcement learning framework integrates advanced techniques such as prioritized learning and experience replay to accelerate the training process and improve learning efficiency. Trained data is saved with a simple string set on the component (in the Inspector), which allows multi agents (or NPC's) to use the same learned data by accessing the same saved path. This enables continuous experimentation and iteration, for developers to freely create trained AI agents for Games/Apps.
Notes:
* There is a single robot model used in the "Dodge Bot" example. This is credited to Quaternius, and is under a CC0 license; see Third-Party Notices.txt file in package for details
* To get the most out of Simple Q, it is advised you read the doc included with the asset
* Use the namespace "QLearning" to access the classes when building! - Technical Details:
- The framework includes the following features:
* Q Learning Algorithm + QTable
* Replay Buffer Experience
* Exploration Vs Exploitation (two methods available)
* Dynamic Decay (two common methods, and a new example method)
* Shared Data Component
* Persistent Data
* Prioritized Learning (with three working examples to choose from)
* Buffer Replay Removal Policy (with four working examples to choose from)
* Attribute Pairs (For better Experience Sampling)
* Little Q (a very basic setup for Q Learning - great for learning the basics)
* Three agents/environments (Virtual Battle Bot, 2D Navigator, Dodge Bot (3D)) - Continue »