PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing

This work introduces PackLab, a comprehensive framework designed to tackle robotic bin packing using multimodal large language models (MLLMs). The system includes PackLab-Suite, a physics-based simulation platform for generating training trajectories; PackLab-VLM, a specialized MLLM that tracks object and container states for closed-loop decision making; and PackLab-Bench, a standardized evaluation suite. Extensive experiments demonstrate that PackLab-VLM outperforms conventional heuristics, traditional reinforcement learning, and general-purpose MLLMs across various configurations.

PackLab-VLM outperforms conventional packing heuristics, traditional reinforcement learning methods, and general-purpose MLLMs across object sets and container configurations.