Highlights:
- New AI language ‘Regular Games’ (RG) introduces an automata-based approach to General Game Playing (GGP).
- The system uses a finite automaton to define game rules for greater computational efficiency.
- RG enables faster simulation models than current GGP platforms like Ludii and Regular Boardgames.
- Developed by Radosław Miernik, Marek Szykuła, Jakub Kowalski, Jakub Cieśluk, Łukasz Galas, and Wojciech Pawlik.
TLDR:
Researchers introduce Regular Games—a novel automata-based language that enhances efficiency and flexibility in General Game Playing systems—surpassing existing frameworks and simplifying both human and AI-driven game design.
A team of computer scientists led by Radosław Miernik, Marek Szykuła, Jakub Kowalski, Jakub Cieśluk, Łukasz Galas, and Wojciech Pawlik has unveiled a groundbreaking approach to Artificial Intelligence in gaming with the introduction of Regular Games (RG), an automata-based General Game Playing (GGP) language. This new system redefines how AI can interpret, design, and simulate complex games by merging computational rigor with design simplicity. Designed for both human creators and automated procedural content generation (PCG), RG sets new performance standards in the field of game-based AI research.
The Regular Games framework is built around a core low-level language powered by finite automata—an elegant mathematical model used to describe state transitions. This minimal yet powerful design ensures that the game rules are easily understandable not only by humans but also by AI agents that perform analysis, optimization, and strategy synthesis. By structuring rules as automata, RG achieves universality for all finite turn-based games with imperfect information, paving the way for advanced AI systems capable of reasoning across a wide variety of game types. The authors demonstrate that forward models built in Regular Games outperform leading GGP systems such as Ludii and Regular Boardgames, delivering superior computational efficiency and faster simulation speeds.
Beyond the theoretical innovation, RG is supported by a rich ecosystem of tools aimed at developers and researchers. These include an integrated editor with Language Server Protocol (LSP) support, automaton visualization, benchmark testing utilities, and a sophisticated debugger that tracks rule transformation from high-level design to low-level automaton representation. This integrated approach enhances reproducibility, debugging, and automated optimization, making it a valuable framework for both AI game research and practical development. As a full version of an upcoming AAAI 2026 paper, this research represents a major leap toward scalable, efficient, and interpretable General Game Playing systems—a milestone for AI-driven game understanding and procedural game design.
The Regular Games project showcases how computational theory and artificial intelligence can jointly address long-standing challenges in GGP, offering a universal platform that balances expressiveness with performance. It exemplifies the ongoing integration of formal automata theory into applied AI research, moving closer to the dream of fully intelligent game-playing systems that can learn, adapt, and compete in a vast range of contexts without human intervention.
Source:
Source:
Original research paper: “Regular Games — an Automata-Based General Game Playing Language” by Radosław Miernik, Marek Szykuła, Jakub Kowalski, Jakub Cieśluk, Łukasz Galas, and Wojciech Pawlik, arXiv:2511.10593v1 [cs.AI], https://arxiv.org/abs/2511.10593