Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]

Optimality Guarantees for Particle Belief Approximation of POMDPs

Authors: Michael H. Lim, Tyler J. Becker, Mykel J. Kochenderfer, Claire J. Tomlin, Zachary N. Sunberg

JAIR 2023 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In addition to our theoretical contribution, we perform five numerical experiments on benchmark POMDPs to demonstrate that a simple MDP algorithm adapted using PB-MDP approximation, Sparse-PFT, achieves performance competitive with other leading continuous observation POMDP solvers.
Researcher Affiliation Academia Michael H. Lim EMAIL University of California, Berkeley, Electrical Engineering and Computer Sciences Department; Tyler J. Becker EMAIL University of Colorado Boulder, Aerospace Engineering Sciences Department; Mykel J. Kochenderfer EMAIL Stanford University, Aeronautics and Astronautics Department; Claire J. Tomlin EMAIL University of California, Berkeley, Electrical Engineering and Computer Sciences Department; Zachary N. Sunberg EMAIL University of Colorado Boulder, Aerospace Engineering Sciences Department
Pseudocode Yes Algorithm 1 Sparse Sampling-ω; Algorithm 2 Sparse-PFT Algorithm
Open Source Code Yes The open source code for the experiments is built on the POMDPs.jl framework (Egorov et al., 2017), and is available at: github.com/Whiffle Fish/PFTExperiments.
Open Datasets Yes we perform five numerical experiments on benchmark POMDPs... The Laser Tag POMDP (Fig. 5) is taken from the DESPOT benchmarks (Ye et al., 2017)
Dataset Splits No The paper describes simulation experiments: 'A total of 5000 simulation experiments were conducted for each configuration combination of solver and environment... and 1000 simulation experiments for the Light Dark and Sub Hunt problems'. This refers to the number of Monte Carlo runs for evaluation, not traditional training/test/validation splits of a dataset.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory, or cloud instance specifications) used for running the experiments. It only mentions planning time limits.
Software Dependencies No The open source code for the experiments is built on the POMDPs.jl framework (Egorov et al., 2017). While POMDPs.jl is a specific framework, its version number is not provided, which is required for reproducibility.
Experiment Setup Yes Appendix F. Experiment Details contains Table 2: Summary of hyperparameters used in experiments. This table lists specific values for c_UCB, beta_UCB, ka, alpha_a, ko, alpha_o, mmin, delta, C, and Depth for various solvers across different environments.