EmbodiedCoder: Parameterized Embodied Mobile Manipulation via Modern Coding Model

2025-10-07

AI Summary

Existing robotic control approaches rely heavily on large-scale annotated datasets, suffering from limited generalizability and poor interpretability. This paper proposes a training-free embodied manipulation framework that enables end-to-end mapping from natural language instructions to robot trajectories via executable code as an intermediary representation. Leveraging large language models (LLMs) to generate interpretable control code—integrated with vision-language perception and motion planning modules—the framework accomplishes mobile manipulation tasks in open-world environments. It eliminates the need for fine-tuning or additional data collection, supports online geometric parameterization of novel objects and real-time trajectory generation, and thereby significantly enhances system transparency, cross-environment generalization, and adaptability to unseen objects and scenarios. Extensive experiments on a real mobile robot demonstrate stable performance across diverse, long-horizon manipulation tasks.

Technology Category

• Intelligent Robots: Embodied AI
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Recent advances in control robot methods, from end-to-end vision-language-action frameworks to modular systems with predefined primitives, have advanced robots'ability to follow natural language instructions. Nonetheless, many approaches still struggle to scale to diverse environments, as they often rely on large annotated datasets and offer limited interpretability.In this work, we introduce EmbodiedCoder, a training-free framework for open-world mobile robot manipulation that leverages coding models to directly generate executable robot trajectories. By grounding high-level instructions in code, EmbodiedCoder enables flexible object geometry parameterization and manipulation trajectory synthesis without additional data collection or fine-tuning.This coding-based paradigm provides a transparent and generalizable way to connect perception with manipulation. Experiments on real mobile robots show that EmbodiedCoder achieves robust performance across diverse long-term tasks and generalizes effectively to novel objects and environments.Our results demonstrate an interpretable approach for bridging high-level reasoning and low-level control, moving beyond fixed primitives toward versatile robot intelligence. See the project page at: https://anonymous.4open.science/w/Embodied-Coder/

Problem

• Generating executable robot trajectories via coding models
• Enabling flexible object geometry parameterization without fine-tuning
• Bridging high-level reasoning with low-level robot control

Innovation

• Leverages coding models to generate robot trajectories
• Enables flexible object geometry parameterization without training
• Provides transparent connection between perception and manipulation