analiza aiqbench
ModuLoop demonstrates that pre-trained LLMs can generate precise, low-level robotic control code without task-specific fine-tuning by employing a closed-loop debugging mechanism. This approach effectively bridges the gap between high-level language models and the real-time, environment-dependent requirements of physical robotic manipulation. The framework's ability to autonomously refine executable programs for tasks like camera calibration and pick-and-place operations highlights a scalable path toward practical LLM-based robotics.
ModuLoop : Low-Level Code Generation using Modular Synthesizer and Closed-Loop Debugger for Robotic Control
arXiv:2606.03047v1 Announce Type: new Abstract: Large Language Models (LLMs) have demonstrated impressive performance across various domains, including code generation and problem solving. However, their application in robotic control, particularly in low-level tasks that require precise manipulation, real-time feedback, and environment-dependent execution, remains limited. To address this challenge, we propose the Closed-Loop Modular Code Synthesizer framework. This framework leverages a pre-trained LLM without any task-specific fine-tuning to perform modular code planning and generation, and iteratively executes the generated code while inserting debugging probes to observe its behavior. This closed-loop structure facilitates systematic debugging and refinement, ultimately producing executable control programs. We apply the proposed framework to the calibration of an RGB-D camera and a robotic arm, validating its effectiveness in real-world settings. Furthermore, through a subsequent pick-and-place task, we demonstrate not only the accuracy of the calibration but also the potential extensibility of the framework. Across both tasks, the framework achieved high execution accuracy and autonomy, illustrating the practicality and scalability of LLM-based robotic control using our framework.
Abstract:Large Language Models (LLMs) have demonstrated impressive performance across various domains, including code generation and problem solving. However, their application in robotic control, particularly in low-level tasks that require precise manipulation, real-time feedback, and environment-dependent execution, remains limited. To address this challenge, we propose the Closed-Loop Modular Code Synthesizer framework. This framework leverages a pre-trained LLM without any task-specific fine-tuning to perform modular code planning and generation, and iteratively executes the generated code while inserting debugging probes to observe its behavior. This closed-loop structure facilitates systematic debugging and refinement, ultimately producing executable control programs. We apply the proposed framework to the calibration of an RGB-D camera and a robotic arm, validating its effectiveness in real-world settings. Furthermore, through a subsequent pick-and-place task, we demonstrate not only the accuracy of the calibration but also the potential extensibility of the framework. Across both tasks, the framework achieved high execution accuracy and autonomy, illustrating the practicality and scalability of LLM-based robotic control using our framework.
Robotics (cs.RO); Multiagent Systems (cs.MA)
arXiv:2606.03047 [cs.RO]
(or arXiv:2606.03047v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2606.03047 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
https://doi.org/10.1109/LRA.2025.3623437 Focus to learn more DOI(s) linking to related resources
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 268 cuvinte
Subiecte conexe
TTCS: Test-Time Curriculum Synthesis for Self-Evolving
arXiv · 20 oct. 2026
Plan-to-Synthesis: Cross-City Human Mobility Generation via Semantic Latent Flow Matching
arXiv · 30 sept. 2026
SketchSong: Hierarchical Song Generation with Sketch Planning and Fine-Grained Multi-Track Modeling
arXiv · 3 iun. 2026
STAB: Specification-driven Testing for Algorithmic Bottlenecks
arXiv · 1 oct. 2026