Preference-Conditioned Reinforcement Learning for Space-Time Efficient Online 3D Bin Packing

2026-03-08

AI Summary

This work addresses the fundamental trade-off in 3D bin packing between spatial efficiency and operational time, where maximizing packing density often conflicts with minimizing execution duration. To resolve this tension, we propose STEP, a reinforcement learning framework based on action selection that, for the first time, incorporates a preference-conditioning mechanism to explicitly model the spatial gain and temporal cost of candidate placement actions. This enables dynamic, time-aware decision-making that balances space utilization and execution efficiency. The approach generalizes across varying candidate set sizes and integrates seamlessly with existing placement modules to yield adaptive, time-sensitive policies. Experimental results demonstrate that STEP reduces operational time by 44% while maintaining packing density at baseline levels.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Search and Optimization: Learning to Search
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Robotic bin packing is widely deployed in warehouse automation, with current systems achieving robust performance through heuristic and learning-based strategies. These systems must balance compact placement with rapid execution, where selecting alternative items or reorienting them can improve space utilization but introduce additional time. We propose a selection-based formulation that explicitly reasons over this trade-off: at each step, the robot evaluates multiple candidate actions, weighing expected packing benefit against estimated operational time. This enables time-aware strategies that selectively accept increased operational time when it yields meaningful spatial improvements. Our method, STEP (Space-Time Efficient Packing), uses a preference-conditioned, Transformer-based reinforcement learning policy, and allows generalization across candidate set sizes and integration with standard placement modules. It achieves a 44% reduction in operational time without compromising packing density. Additional material is available at https://step-packing.github.io.

Problem

• 3D bin packing
• space-time trade-off
• robotic automation
• operational efficiency
• online packing

Innovation

• Preference-Conditioned Reinforcement Learning
• Space-Time Trade-off
• Online 3D Bin Packing
• Transformer-based Policy
• Warehouse Automation