Stratified Strategy Selection for Unit Control in Real-Time Strategy Games

Authors: Levi H. S. Lelis

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results on a simulator of an RTS game shows that SSS employing either fixed or adaptive type systems is able to substantially outperform stateof-the-art search-based algorithms in combat scenarios with up to 100 units.
Researcher Affiliation Academia Levi H. S. Lelis Departamento de Inform atica, Universidade Federal de Vic osa, Brazil levi.lelis@ufv.br
Pseudocode Yes Algorithm 1 Stratified Strategy Selection
Open Source Code No The paper mentions 'Spar Craft codebase [Churchill and Buro, 2013]' and provides a link to 'github.com/davechurchill/ualbertabot/tree/master/Spar Craft', which refers to a third-party simulation environment used in the paper, not the source code for the proposed SSS/SSS+ methods themselves. There is no explicit statement or link indicating the release of the authors' own implementation code.
Open Datasets Yes Our experiments are run in Spar Craft, a simplified combat simulation environment of Blizzard s Star Craft [Churchill and Buro, 2013]. For each combat scenario we generate 1,000 start states as explained by Churchill and Buro [2013].
Dataset Splits No The paper describes testing algorithms on 1,000 generated start states for various combat scenarios but does not specify traditional train/validation/test dataset splits with percentages, sample counts, or explicit splitting methodologies for reproducibility.
Hardware Specification No All experiments are run on 2.66 GHz machines. This is a general clock speed but lacks specific details such as CPU model, GPU, or memory for hardware specification.
Software Dependencies No The paper mentions using 'Spar Craft' as the simulation environment and implementing 'POE in Spar Craft s codebase', but it does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes All algorithms use the same set Σ = {NOKAV, Kiter, Cluster} of scripts, and have a time limit of 40 milliseconds for each decision point. For SSS we use Tc,3 and for SSS+ we use Y = {Tc,3, Tc,2, Tc,1, Tc,0, TRGD}... PGS s and PGS+ s improvement response parameter to 0... POE: population size of 36, with the 6 fittest individuals being selected for generating 5 offsprings each. Also, POE uses a playout-based evaluation function that is limited to 25 decision points.