Online automatic code generation for robot swarms: LLMs and self-organizing hierarchy

2025-10-06

AI Summary

To address the lack of online behavioral adaptability in robot swarms when encountering obstacles during task execution, this paper proposes an online autonomous code generation framework integrating Large Language Models (LLMs) with the Self-Organizing Networked Systems (SoNS) paradigm. Methodologically, we pioneer the embedding of LLMs into a distributed hierarchical control architecture, enabling real-time obstacle perception, autonomous code request initiation, and on-the-fly generation and deployment of executable behavior scripts; SoNS underpins environmental sensing, task decomposition, and swarm-level coordination. Evaluated on six physical robots and simulations with over 30 nodes, the approach achieves an 85% task recovery success rate. Results demonstrate substantial improvements in dynamic adaptability, scalability, and engineering feasibility. The core contribution is the first realization of LLM-driven, closed-loop online code generation and execution for robot swarms.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Our recently introduced self-organizing nervous system (SoNS) provides robot swarms with 1) ease of behavior design and 2) global estimation of the swarm configuration and its collective environment, facilitating the implementation of online automatic code generation for robot swarms. In a demonstration with 6 real robots and simulation trials with >30 robots, we show that when a SoNS-enhanced robot swarm gets stuck, it can automatically solicit and run code generated by an external LLM on the fly, completing its mission with an 85% success rate.

Problem

• Online automatic code generation for robot swarms
• Self-organizing hierarchy enables global swarm configuration
• LLMs generate recovery code when swarm gets stuck

Innovation

• Self-organizing nervous system for robot swarms
• Online automatic code generation using LLMs
• Dynamic code solicitation during mission execution