NeSyPack: A Neuro-Symbolic Framework for Bimanual Logistics Packing

2025/06/06 by Bowei Li, Li, Bowei, Peiqi Yu +11
Engineering · #Robot Manipulation and Learning #Modular Robots and Swarm Intelligence #Flexible and Reconfigurable Manufacturing Systems

paper · pdf · doi:10.48550/arxiv.2506.06567

Abstract

This paper presents NeSyPack, a neuro-symbolic framework for bimanual logistics packing. NeSyPack combines data-driven models and symbolic reasoning to build an explainable hierarchical system that is generalizable, data-efficient, and reliable. It decomposes a task into subtasks via hierarchical reasoning, and further into atomic skills managed by a symbolic skill graph. The graph selects skill parameters, robot configurations, and task-specific control strategies for execution. This modular design enables robustness, adaptability, and efficient reuse - outperforming end-to-end models that require large-scale retraining. Using NeSyPack, our team won the First Prize in the What Bimanuals Can Do (WBCD) competition at the 2025 IEEE International Conference on Robotics and Automation.

Citations

• Stow: Robotic Packing of Items into Fabric Pods
• π_0.5: a Vision-Language-Action Model with Open-World Generalization
• APEX-MR: Multi-Robot Asynchronous Planning and Execution for Cooperative Assembly
• Physics-Aware Robotic Palletization with Online Masking Inference
• π₀: A Vision-Language-Action Flow Model for General Robot Control
• Automating Robot Failure Recovery Using Vision-Language Models With Optimized Prompts
• PaliGemma: A versatile 3B VLM for transfer
• OpenVLA: An Open-Source Vision-Language-Action Model
• Decomposition-based Hierarchical Task Allocation and Planning for Multi-Robots under Hierarchical Temporal Logic Specifications
• Llama 2: Open Foundation and Fine-Tuned Chat Models
• DINOv2: Learning Robust Visual Features without Supervision
• Sigmoid Loss for Language Image Pre-Training
• Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding
• The Convex Feasible Set Algorithm for Real Time Optimization in Motion Planning
• The Convex Feasible Set Algorithm for Real Time Optimization in Motion Planning

Related

• Learning Neuro-Symbolic Skills for Bilevel Planning
• Learning Neuro-Symbolic Relational Transition Models for Bilevel Planning
• Fast and Accurate Task Planning using Neuro-Symbolic Language Models and Multi-level Goal Decomposition
• NeSyPr: Neurosymbolic Proceduralization For Efficient Embodied Reasoning
• VisualPredicator: Learning Abstract World Models with Neuro-Symbolic Predicates for Robot Planning