Learning to Staff: Offline Reinforcement Learning and Fine-Tuned LLMs for Warehouse Staffing Optimization

TL;DR

This paper explores offline reinforcement learning and fine-tuned large language models to optimize warehouse staffing, demonstrating improvements in throughput and decision-making support in simulated environments.

Contribution

It introduces two novel approaches—Transformer-based offline RL policies and fine-tuned LLMs—for warehouse staffing optimization, comparing their effectiveness and practical applicability.

Findings

Offline RL achieved 2.4% throughput improvement.

Fine-tuned LLMs matched or exceeded baseline performance.

Prompting alone was insufficient for effective decision-making.

Abstract

We investigate machine learning approaches for optimizing real-time staffing decisions in semi-automated warehouse sortation systems. Operational decision-making can be supported at different levels of abstraction, with different trade-offs. We evaluate two approaches, each in a matching simulation environment. First, we train custom Transformer-based policies using offline reinforcement learning on detailed historical state representations, achieving a 2.4% throughput improvement over historical baselines in learned simulators. In high-volume warehouse operations, improvements of this size translate to significant savings. Second, we explore LLMs operating on abstracted, human-readable state descriptions. These are a natural fit for decisions that warehouse managers make using high-level operational summaries. We systematically compare prompting techniques, automatic prompt optimization, and fine-tuning strategies. While prompting alone proves insufficient, supervised fine-tuning combined with Direct Preference Optimization on simulator-generated preferences achieves performance that matches or slightly exceeds historical baselines in a hand-crafted simulator. Our findings demonstrate that both approaches offer viable paths toward AI-assisted operational decision-making. Offline RL excels with task-specific architectures. LLMs support human-readable inputs and can be combined with an iterative feedback loop that can incorporate manager preferences.