When Search Becomes Memory: Turning Robot Design Trials into Transferable Skills
2026-05-25 • Robotics
RoboticsArtificial IntelligenceComputation and LanguageComputer Vision and Pattern Recognition
AI summaryⓘ
The authors developed Auto-Robotist, a system where a language model learns and remembers design rules for robot bodies during evolutionary experiments. Instead of starting fresh each time, it creates a library of skills written in natural language that capture what works and what doesn't, making the design process easier to understand and reuse. This skill library helps improve robot designs and allows knowledge to transfer to bigger robot design tasks more effectively than traditional methods. Their tests show this approach leads to better initial designs and successful learning across different robot challenges.
large language modelsevolutionary robot designgenetic algorithmsmorphology searchskill librarynatural language processingdesign memorytransfer learningrobot locomotionsimulator evaluation
Authors
Yunfei Wang, Xiaohao Xu, Yang Li, Xiaonan Huang
Abstract
Large language models (LLMs) are increasingly used as proposal generators for evolutionary robot design, yet most loops remain memoryless: simulator results shape the next population but are not preserved as reusable design knowledge. We present Auto-Robotist, a self-evolving LLM agent that distills morphology-search traces into an explicit natural-language skill library. Each skill stores a structural archetype, evidence-grounded positive and negative rules, and the evaluated designs that support them, making design memory inspectable rather than implicit in a population. During search, the agent retrieves skills to condition LLM edits of elite bodies while retaining a Genetic Algorithm (GA) mutation path for exploration; after evaluation, it updates the library through Add, Diagnose, and Merge. Across seven EvoGym tasks spanning locomotion, traversal, and object interaction, Auto-Robotist improves cold-start 5x5 search and transfers learned skills to 10x10 design spaces, where reference-conditioned transfer outperforms GA on every task. These results suggest that LLM agents can convert expensive physical evaluations into reusable, auditable design principles. Our code will be released upon acceptance.