Evolving Action Abstractions for Real-Time Planning in Extensive-Form Games

Authors: Julian R. H. Mariño, Rubens O. Moraes, Claudio Toledo, Levi H. S. Lelis2330-2337

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results on small matches of µRTS show that our evolutionary approach is able to converge to a Nash equilibrium for the subset selection game. Also, results on larger matches show that search algorithms using action abstractions derived by our evolutionary approach are able to substantially outperform all state-of-the-art planning systems tested.
Researcher Affiliation Academia Julian R. H. Mari no,1 Rubens O. Moraes,2 Claudio Toledo,1 Levi H. S. Lelis2 1Departamento de Sistemas de Computac ao, ICMC, Universidade de S ao Paulo, Brazil 2Departamento de Inform atica, Universidade Federal de Vic osa, Brazil
Pseudocode Yes Algorithm 1 Evolutionary Algorithm for Solving SSG
Open Source Code Yes See our codebase for details of how the 300 strategies are generated: https://github.com/julianmarino/ evolutionary-action-abstractions.git
Open Datasets No All our experiments are run on µRTS, an RTS game developed for research (Onta n on 2013). The paper uses a game environment for experiments, not a pre-existing publicly available dataset in the traditional sense, nor does it provide a link to the game's data.
Dataset Splits No The paper does not mention specific training, validation, and test dataset splits for reproducibility. It describes an evolutionary algorithm for generating action abstractions and then evaluates their performance in simulated game matches.
Hardware Specification No The paper does not provide specific hardware specifications (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions the µRTS game platform and various algorithms used, but does not list specific software dependencies with version numbers for reproducibility (e.g., programming languages, libraries, or solvers with their versions).
Experiment Setup Yes In the first experiment we use a mutation probability u of 0.05, size of the population n of 100, the number of generations l of 20, and the size e of the elite population is set to 25. Since the maps are much larger in the second experiment, we set n = 20, l = 30, and e = 5.