A hybrid fuzzy logic/constraint satisfaction problem approach to automatic decision making in simulation game models

Area of Science:

• Artificial Intelligence
• Computational Game Theory
• Simulation Modeling

Background:

• Developing AI decision-making that mimics human experts in complex simulations is challenging.
• Existing methods may not adequately capture nuanced decision processes in dynamic environments.
• Military headquarters decision-making provides a relevant model for complex strategic choices.

Purpose of the Study:

• To design and evaluate a hybrid fuzzy logic (FL) and constraint satisfaction problem (CSP) decision agent for simulation games.
• To test the general applicability of the FL/CSP agent in diverse simulation environments.
• To demonstrate the agent's ability to approximate human expert decision-making.

Main Methods:

• A hybrid approach combining fuzzy logic (FL) and constraint satisfaction problem (CSP) for decision agent design.
• Modeling military headquarters decision processes to inform agent variable selection and rulebases.
• Application and testing of the FL/CSP agent in two distinct simulation games: an air campaign game and a network flow stochastic board game.

Main Results:

• The hybrid FL/CSP decision agent demonstrated effective performance in both tested simulation games.
• The design proved generally applicable, even in a complex board game with a vast action set.
• Successful training of automatic FL/CSP decision agents against defined performance measures was achieved.

Conclusions:

• The hybrid fuzzy logic and constraint satisfaction problem approach offers a robust method for automatic decision-making in complex simulation models.
• This design effectively approximates human expert behavior, showing promise for advanced AI in strategic simulations.
• Further research can explore expanded applications and refinements of the FL/CSP agent training and performance.