Sampling Networks and Aggregate Simulation for Online POMDP Planning

Authors: Hao(Jackson) Cui, Roni Khardon

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate SNAP on problems from the international planning competition (IPC) 2011, the latest IPC with publicly available challenge POMDP problems, comparing its performance to POMCP [21] and DESPOT [25]. The results show that the algorithm is competitive on a large range of problems and that it has a significant advantage on large problems. 6 Experimental evaluation We compare the performance of SNAP to the state-of-the-art online planners for POMDP.
Researcher Affiliation Academia Hao Cui Department of Computer Science Tufts University Medford, MA 02155, USA hao.cui@tufts.edu Roni Khardon Department of Computer Science Indiana University Bloomington, IN, USA rkhardon@iu.edu
Pseudocode No The paper does not contain structured pseudocode or clearly labeled algorithm blocks; procedures are described in narrative text and mathematical equations.
Open Source Code No The paper mentions using third-party implementations for DESPOT and POMCP but does not provide any statement or link for the open-source code of their own SNAP algorithm.
Open Datasets Yes We evaluate SNAP on problems from the international planning competition (IPC) 2011, the latest IPC with publicly available challenge POMDP problems, comparing its performance to POMCP [21] and DESPOT [25].
Dataset Splits No The paper evaluates performance on predefined IPC problem instances and does not describe dataset splits (e.g., training, validation, test) for reproducibility.
Hardware Specification No This work was partly supported by NSF under grant IIS-1616280 and by an Adobe Data Science Research Award. Some of the experiments in this paper were performed on the Tufts Linux Research Cluster supported by Tufts Technology Services.
Software Dependencies No The paper mentions using the RDDL language and refers to implementations of other algorithms (DESPOT, POMCP) but does not provide specific version numbers for software dependencies used in their own work.
Experiment Setup Yes For the main experiment we use 2 seconds planning time per step for all planners. In particular we fix the search depth (to 5) and the number of updates (to 200) and repeat the experiment 100 times.