Project Blitzkrieg

Project Blitzkrieg

Distributed, scalable and extremely easy to use game decision engine based on Monte-Carlo Tree Search algorithm.

Get the code Read the documentation

Rapid setup

The setup of Blitzkrieg solution should not take longer than 20 minutes.

Scalable by design

You can adjust the computing power of Blitzkrieg on the fly.

Easy to use

All functions are wrapped in modern webpage UI and well-designed HTTP API.

 

One could think of Blitzkrieg project as “Artificial Intelligence for games As a Service” – the intention is to create ease to use, working out-of-the-box solution that will predict best possible moves for any game it is provided with. Moreover, it will be written with scalability in mind, so that it could adjust to the quantity of requests it is receiving or increase its prediction accuracy on the fly, without any effort.

As it is impossible to create game decision making algorithm that plays perfectly every possible game, the responses from Blitzkrieg will be not guaranteed to be best possible. However, they can be considered to be in reasonable boundaries of “good moves”. This will be achieved due to usage of distributed, parallel Monte Carlo Tree Search as a heart of the decision engine – the algorithm does not need to know theory of the game it is playing, it only needs to know how to play it.

The envisioned user-experience is as follows: customer downloads Blitzkrieg projects and deploys it to cloud provider such as Azure or AWS. Then it supplies the service with implementation of the game he wishes to predict moves for. After that, usage is as simple ask asking for best moves in given situation via HTTP API.

Where to next ?

You can learn more about the project reading introduction page.

You can fork the con on our GitHub Repository.