Solving Partially Observable Stochastic Games with Public Observations

Authors: Karel Horák, Branislav Bošanský2029-2036

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that our algorithm can closely approximate the value of non-trivial games with hundreds of states.
Researcher Affiliation Academia Karel Hor ak, Branislav Boˇsansk y Department of Computer Science, Faculty of Electrical Engineering Czech Technical University in Prague {horak,bosansky}@agents.fel.cvut.cz
Pseudocode Yes Algorithm 1: HSVI algorithm for PO-POSGs.
Open Source Code No The paper does not provide any statement or link indicating that its source code is open or publicly available.
Open Datasets No The paper describes generating instances for the Patrolling and Lasertag domains (e.g., 'generate random graphs from the Dorogovtsev Mendes model', 'randomly placed obstacles') rather than using pre-existing, publicly available datasets for which access information is provided.
Dataset Splits No The paper describes the construction of game instances and parameters for experiments but does not specify dataset splits (e.g., training, validation, test percentages or counts) in the typical machine learning sense.
Hardware Specification Yes All experiments use discount factor γ = 0.95 and were run on Intel i7-8700K (solving 6 instances in parallel).
Software Dependencies No The paper discusses the use of linear programming and implies the use of a solver but does not specify any software dependencies with version numbers.
Experiment Setup Yes All experiments use discount factor γ = 0.95... In our experimental evaluation, we consider t = 3 and t = 4... We ran the algorithm with ϵ = 0.05 for 5 hours.