Advancing AI in Video Games with AMD Schola

Introducing AMD Schola

What if you could integrate cutting-edge reinforcement learning (RL) algorithms directly into your gaming projects - without lengthy workarounds or rebuilding environments from scratch?

AMD Schola is here to make that a reality. By connecting popular open-source RL libraries (written in Python) with the visual and physics capabilities of Unreal Engine, Schola empowers AI researchers and game developers alike to push the boundaries of intelligent gameplay.

The Challenge: Bridging Research and Game Development

AI researchers often develop sophisticated RL algorithms in isolated simulation environments or libraries, only to encounter challenges porting these solutions into actual games.

Meanwhile, game developers aiming to add intelligent behaviors to non-player characters (NPCs) usually must reinvent the wheel - creating new AI logic from scratch or attempting complex integrations with minimal documentation.

Why does this matter?

• It slows down innovation: promising research remains confined to academia.

• It adds complexity: developers must replicate algorithms or shoehorn models trained elsewhere into existing game mechanics, leading to suboptimal results.

A Stealth Game Example

Imagine a stealth game with NPC guards, currently developers must strike a difficult balance between having guards always know exactly where the player is (making the game difficult, and unrealistic) or have little awareness at all (making it too easy).

With Schola, these guards can learn realistic behaviors - like methodically searching for the player or intelligently responding to suspicious events. This flexible toolkit for AI development takes gameplay from predictable to immersive.

Envisioning Smarter Gameplay

For Game Developers

• More engaging NPCs: NPCs adapt to player actions in real time, leading to dynamic gameplay that keeps players engaged.

• Faster prototyping: With Schola, you can experiment with a variety of RL-driven behaviors, testing them quickly in the Unreal Editor.

For AI Researchers

• Direct Unreal integration: Take advantage of Unreal Engine’s advanced physics, rendering, and asset libraries.

• Streamlined workflows: No more hacking together prototypes and porting them. Train, test, and refine RL algorithms in the same environment you’ll use for final deployment.

The AMD Schola Advantage

Schola supports two popular open-source RL libraries (RLlib and Stable-Baselines3) and the Gymnasium API, making it highly accessible for anyone already familiar with these tools. It provides:

• Multiple RL library integrations: Seamlessly switch between or combine your favorite RL frameworks without learning a new API.

• Immediate Unreal Engine compatibility: All training, testing, and debugging can happen directly within Unreal’s rich ecosystem - no environment rebuilding necessary.

• Easy setup and documentation: Our online API docs walk you through the process of installing and using Schola, so you can start training your models faster.

Showcasing the Tech

If you want a sneak peek of Schola in action, check out our examples. You’ll see how RL models can create more natural behaviors and actions.

The technical design diagram below details how Schola communicates with both Unreal Engine and the RL libraries.

Getting Started and Taking Action

Ready to dive in? We offer several ways to explore Schola, each tailored to fit different levels of interest or expertise:

1. Install the plugin

   • Bring Schola into an existing Unreal project to start experimenting with RL-driven NPCs right away.

2. Try our example environments

   • Download examples from GPUOpen on GitHub - such as MazeSolver, Tag, Pong, and BallShooter - to see RL-driven gameplay in action inside the Unreal Editor.

3. Explore our documentation

   • Check out our official docs here on GPUOpen, for step-by-step tutorials on building custom environments from scratch, and to learn more about advanced features in Python and Unreal.

4. Contribute and engage

   • Visit our public GitHub repository for Schola to access source code, report bugs, and request new features. We thrive on community feedback and welcome pull requests.

Looking Ahead

By merging the strengths of open-source RL tools and Unreal Engine, Schola opens new doors for next-level AI in gaming. Researchers can focus on refining algorithms, while developers can easily implement them for more immersive and adaptive gameplay.

Stay tuned for new features, guides, and examples to keep advancing your ability to build immersive AI!

Ready to see what AMD Schola can do for your projects?

• Install the plugins, explore the examples, and witness the difference that a learning-based approach can make.

• Join the community on GitHub and help shape the future of AI-driven NPCs.

We can’t wait to see the groundbreaking ideas and intelligent game worlds you’ll create with Schola.

Useful links

AMD Schola home page	AMD Schola documentation here on GPUOpen