The project concerns creating a component-based engine for embedding advanced Artificial Intelligence (AI) in games. The main usage of the AI will be to equip actors or group of actors (often also referred to as characters) with realistic sophiscated behavior. The actors will be able to eﬃciently drive towards their individually set goals (e.g. trying to beat the human opponent in a warfare game) as well as adapt to the human player’s way of playing. The technological solution, which enables realizations of all the final results, can be regarded as an intermediate result, crucial to successful meeting of the goals set. The technological core is a complex AI system based on two major methods: Monte Carlo Tree Search (to search the game state-space and perform action reasoning) and Deep Neural Networks (to learn specific features in games and provide a classier of game states). Such an integration requires to solve certain technological issues, therefore, we will likely contribute to the stateof-the-art in the areas of MCTS, DNN and game playing.
The outcome of the project will be a software module, which, at the time of submiting the present application, has no equivalent on the market. It could be used in the production of games in order to automatize certain stages of character creation, related to the automatic control of their behavior. Such level of automation is unique on the market. The project provides very powerful Artificial Intelligence, which only requires the knowledge about a game’s world and the possibility to evaluate its actions in order to be able to improve its own eﬀectiveness in time. Our solution will make it possible to free resources that until now have been needed for manual improvement of soti methods and applying these resources to a creative use of a wide spectrum of playing skills.
Beneficiary: Silver Bullet Labs Sp. z o.o.
Value of the project: 5,571,400 PLN
EU donation: 4,155,095.65 PLN