Learning Efficient and Fair Policies for Uncertainty-Aware Collaborative Human-Robot Order Picking

TL;DR

This work tackles online, uncertainty-aware allocation of human pickers to autonomous mobile robots in warehouses by formulating a dual-objective optimization: minimize total completion time $C$ and reduce workload fairness disparity $\sigma(W_1, dots,W_{| obreak K|})$. It introduces DRL-Guided Picker Optimization, combining a discrete-event simulation with a graph-based state representation and the novel Aisle-Embedding Multi-Objective Aware Network (AEMO-Net) to capture regional warehouse structure and separate efficiency from fairness signals. A multi-objective learning algorithm builds a Pareto front of non-dominated policies using a warm-up population, predictive rewards, and PPO-based optimization, demonstrating strong transferability to different warehouse sizes and picker/AMR ratios. Empirical results show policies surpass greedy and VI baselines in both efficiency and fairness, with substantial improvements in large warehouses, and practical transferability that supports industrial deployment and broader applicability to real-world uncertainty in collaborative order picking.

Abstract

In collaborative human-robot order picking systems, human pickers and Autonomous Mobile Robots (AMRs) travel independently through a warehouse and meet at pick locations where pickers load items onto the AMRs. In this paper, we consider an optimization problem in such systems where we allocate pickers to AMRs in a stochastic environment. We propose a novel multi-objective Deep Reinforcement Learning (DRL) approach to learn effective allocation policies to maximize pick efficiency while also aiming to improve workload fairness amongst human pickers. In our approach, we model the warehouse states using a graph, and define a neural network architecture that captures regional information and effectively extracts representations related to efficiency and workload. We develop a discrete-event simulation model, which we use to train and evaluate the proposed DRL approach. In the experiments, we demonstrate that our approach can find non-dominated policy sets that outline good trade-offs between fairness and efficiency objectives. The trained policies outperform the benchmarks in terms of both efficiency and fairness. Moreover, they show good transferability properties when tested on scenarios with different warehouse sizes. The implementation of the simulation model, proposed approach, and experiments are published.

Figures (9)

• Figure 1: Overview of DRL-Guided Picker Optimization.
• Figure 2: Illustration of the considered warehouse parameters and the associated undirected graph representation of a warehouse. The dotted arrows represent the allowed amr movement directions, and the grey areas represent storage racks between the aisles. The circles indicate nodes and the connections between the circles are edges.
• Figure 3: Illustration of the AEMO-Net architecture.
• Figure 4: Performance of the non-dominated sets of policies learned on warehouse different warehouse types.
• Figure 5: Performance of the policies learned on warehouse type S when evaluated on different picker/AMR numbers.