Tighter Value-Function Approximations for POMDPs
(2025) 24th International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2025 In Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS p.1200-1208- Abstract
Solving partially observable Markov decision processes (POMDPs) typically requires reasoning about the values of exponentially many state beliefs. Towards practical performance, state-of-the-art solvers use value bounds to guide this reasoning. However, sound upper value bounds are often computationally expensive to compute, and there is a tradeoff between the tightness of such bounds and their computational cost. This paper introduces new and provably tighter upper value bounds than the commonly used fast informed bound. Our empirical evaluation shows that, despite their additional computational overhead, the new upper bounds accelerate state-of-the-art POMDP solvers on a wide range of benchmarks.
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/c4edf905-ccf8-4c0b-99b1-1e9d2e0baac0
- author
- Krale, Merlijn
; Koops, Wietze
LU
; Junges, Sebastian
; Simão, Thiago D.
and Jansen, Nils
- organization
- publishing date
- 2025
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- Heuristic Search, Planning, POMDPs, Value Bounds
- host publication
- Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2025
- series title
- Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS
- editor
- Vorobeychik, Yevgeniy ; Das, Sanmay and Nowe, Ann
- pages
- 9 pages
- publisher
- International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)
- conference name
- 24th International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2025
- conference location
- Detroit, United States
- conference dates
- 2025-05-19 - 2025-05-23
- external identifiers
-
- scopus:105009828484
- ISSN
- 1558-2914
- 1548-8403
- ISBN
- 9798400714269
- language
- English
- LU publication?
- yes
- additional info
- Publisher Copyright: © 2025 International Foundation for Autonomous Agents and Multiagent Systems (www.ifaamas.org).
- id
- c4edf905-ccf8-4c0b-99b1-1e9d2e0baac0
- date added to LUP
- 2026-01-20 16:13:10
- date last changed
- 2026-01-21 03:49:07
@inproceedings{c4edf905-ccf8-4c0b-99b1-1e9d2e0baac0,
abstract = {{<p>Solving partially observable Markov decision processes (POMDPs) typically requires reasoning about the values of exponentially many state beliefs. Towards practical performance, state-of-the-art solvers use value bounds to guide this reasoning. However, sound upper value bounds are often computationally expensive to compute, and there is a tradeoff between the tightness of such bounds and their computational cost. This paper introduces new and provably tighter upper value bounds than the commonly used fast informed bound. Our empirical evaluation shows that, despite their additional computational overhead, the new upper bounds accelerate state-of-the-art POMDP solvers on a wide range of benchmarks.</p>}},
author = {{Krale, Merlijn and Koops, Wietze and Junges, Sebastian and Simão, Thiago D. and Jansen, Nils}},
booktitle = {{Proceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2025}},
editor = {{Vorobeychik, Yevgeniy and Das, Sanmay and Nowe, Ann}},
isbn = {{9798400714269}},
issn = {{1558-2914}},
keywords = {{Heuristic Search; Planning; POMDPs; Value Bounds}},
language = {{eng}},
pages = {{1200--1208}},
publisher = {{International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS)}},
series = {{Proceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS}},
title = {{Tighter Value-Function Approximations for POMDPs}},
year = {{2025}},
}