Physics-Aware Robotic Palletization with Online Masking Inference

2025-02-19

AI Summary

Existing online palletizing methods struggle with dynamically arriving, irregular boxes exhibiting heterogeneous physical properties (e.g., density, stiffness), as they neglect realistic physical constraints—leading to insufficient stacking stability. This paper proposes a physics-aware online palletizing planning framework. It introduces, for the first time, an online-learning-based action-space masking mechanism that replaces hand-crafted heuristic stability criteria, enabling adaptive, real-time physical constraint reasoning. The framework integrates reinforcement learning with multi-physics modeling—including mass distribution, contact mechanics, and deformation behavior—and is validated in both high-fidelity simulation and on a real-world warehouse robotic platform. Experiments demonstrate significant improvements in stacking success rate and robustness over state-of-the-art methods. Furthermore, industrial-scale deployment confirms its practical feasibility and scalability.

Technology Category

• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Intelligent Robots: Motion and Path Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

The efficient planning of stacking boxes, especially in the online setting where the sequence of item arrivals is unpredictable, remains a critical challenge in modern warehouse and logistics management. Existing solutions often address box size variations, but overlook their intrinsic and physical properties, such as density and rigidity, which are crucial for real-world applications. We use reinforcement learning (RL) to solve this problem by employing action space masking to direct the RL policy toward valid actions. Unlike previous methods that rely on heuristic stability assessments which are difficult to assess in physical scenarios, our framework utilizes online learning to dynamically train the action space mask, eliminating the need for manual heuristic design. Extensive experiments demonstrate that our proposed method outperforms existing state-of-the-arts. Furthermore, we deploy our learned task planner in a real-world robotic palletizer, validating its practical applicability in operational settings.

Problem

• Online planning for unpredictable box arrivals
• Incorporating physical properties in stacking decisions
• Dynamic training of action space masks

Innovation

• Reinforcement learning for palletization
• Online action space masking
• Dynamic training of RL policy