# Stick-Breaking Policy Learning in Dec-POMDPs

Source: [Stick-Breaking Policy Learning in Dec-POMDPs](<https://www.sota2.com/research/paper/68fa622b1e209131ce9f7d42>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-05T05:00:40.919Z

Authors: Miao Liu, Christopher Amato, Xuejun Liao, Lawrence Carin, Jonathan P. How

Published: 2015-05-01T20:29:27.000Z

## Abstract

Expectation maximization (EM) has recently been shown to be an efficient algorithm for learning finite-state controllers (FSCs) in large decentralized POMDPs (Dec-POMDPs). However, current methods use fixed-size FSCs and often converge to maxima that are far from optimal. This paper considers a variable-size FSC to represent the local policy of each agent. These variable-size FSCs are constructed using a stick-breaking prior, leading to a new framework called \\emph{decentralized stick-breaking policy representation} (Dec-SBPR). This approach learns the controller parameters with a variational Bayesian algorithm without having to assume that the Dec-POMDP model is available. The performance of Dec-SBPR is demonstrated on several benchmark problems, showing that the algorithm scales to large problems while outperforming other state-of-the-art methods.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Dec-POMDP Policy Learning and Planning on BROADCAST (4, 2, 5)](<https://www.sota2.com/research/sota/dec-pomdp-policy-learning-and-planning-on-broadcast-4-2-5.md>): Dec-POMDP Policy Learning and Planning; BROADCAST (4, 2, 5); Policy Value: 9.27.

- [Dec-POMDP Policy Learning and Planning on MARS ROVERS 256, 6, 8](<https://www.sota2.com/research/sota/dec-pomdp-policy-learning-and-planning-on-mars-rovers-256-6-8.md>): Dec-POMDP Policy Learning and Planning; MARS ROVERS 256, 6, 8; Policy Value: 20.62.

- [Dec-POMDP Policy Learning and Planning on DEC-TIGER 2, 3, 3](<https://www.sota2.com/research/sota/dec-pomdp-policy-learning-and-planning-on-dec-tiger-2-3-3.md>): Dec-POMDP Policy Learning and Planning; DEC-TIGER 2, 3, 3; Policy Value: -18.63.

- [Dec-POMDP Policy Learning and Planning on RECYCLING ROBOTS (3, 3, 2)](<https://www.sota2.com/research/sota/dec-pomdp-policy-learning-and-planning-on-recycling-robots-3-3-2.md>): Dec-POMDP Policy Learning and Planning; RECYCLING ROBOTS (3, 3, 2); Policy Value: 31.26.

- [Dec-POMDP Policy Learning and Planning on BOX PUSHING 100, 4, 5](<https://www.sota2.com/research/sota/dec-pomdp-policy-learning-and-planning-on-box-pushing-100-4-5.md>): Dec-POMDP Policy Learning and Planning; BOX PUSHING 100, 4, 5; Policy Value: 77.65.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1505.00274>)

[PDF](<https://arxiv.org/pdf/1505.00274.pdf>)