# Memory Bounded Open-Loop Planning in Large POMDPs using Thompson Sampling

## TL;DR

This paper introduces POSTS, a memory-bounded open-loop planning method for large POMDPs using Thompson Sampling, offering a competitive performance-memory tradeoff suitable for resource-constrained environments.

## Contribution

The paper proposes POSTS, a novel memory-bounded open-loop planning algorithm using Thompson Sampling for large POMDPs, addressing memory limitations of existing tree-based methods.

## Key findings

- POSTS performs competitively with tree-based approaches.
- POSTS provides a favorable performance-memory tradeoff.
- Empirical results on benchmark problems validate POSTS' effectiveness.

## Abstract

State-of-the-art approaches to partially observable planning like POMCP are based on stochastic tree search. While these approaches are computationally efficient, they may still construct search trees of considerable size, which could limit the performance due to restricted memory resources. In this paper, we propose Partially Observable Stacked Thompson Sampling (POSTS), a memory bounded approach to open-loop planning in large POMDPs, which optimizes a fixed size stack of Thompson Sampling bandits. We empirically evaluate POSTS in four large benchmark problems and compare its performance with different tree-based approaches. We show that POSTS achieves competitive performance compared to tree-based open-loop planning and offers a performance-memory tradeoff, making it suitable for partially observable planning with highly restricted computational and memory resources.