Game AI Research with Fast Planet Wars Variants [link]
This paper introduces a fast variant of the Planet Wars strategy game specifically tailored for Game AI research, enabling fast trials and agent evaluations, customizable rules, and forward-planning support.
Efficient Evolutionary Methods for Game Agent Optimisation [link]
This paper investigates the efficiency of evolutionary methods for optimizing game agents in Planet Wars. It compares model-based approaches, such as the N-Tuple Bandit Evolutionary Algorithm and Sequential Model-based Algorithm Configuration (SMAC), against non-model-based algorithms, showing that model-based methods perform better in stochastic game environments
Designing Competitive Bots for a RTS Game using Genetic Programming [link]
This paper proposes using Genetic Programming to generate decision-tree-based AI agents for Planet Wars. It addresses the challenge of "noisy" fitness evaluations in dynamic game environments by comparing three different fitness functions: one based on number of victories, and two based on game-metrics. The study shows that while all generated agents outperformed a baseline bot, the win-based fitness function consistently produced the most competitive agents.