Efficient Robot Design with Multi-Objective Black-Box Optimization and Large Language Models

Various methods for robot design optimization have been developed so far. These methods are diverse, ranging from numerical optimization to black-box optimization. While numerical optimization is fast, it is not suitable for cases involving complex structures or discrete values, leading to frequent use of black-box optimization instead. However, black-box optimization suffers from low sampling efficiency and takes considerable sampling iterations to obtain good solutions. In this study, we propose a method to enhance the efficiency of robot body design based on black-box optimization by utilizing large language models (LLMs). In parallel with the sampling process based on black-box optimization, sampling is performed using LLMs, which are provided with problem settings and extensive feedback. We demonstrate that this method enables more efficient exploration of design solutions and discuss its characteristics and limitations.

💡 Research Summary

**
This paper addresses the challenge of efficiently optimizing robot body designs that involve both continuous and discrete variables while simultaneously minimizing multiple objectives. Traditional black‑box optimization (BBO) methods such as genetic algorithms are widely used because they can handle complex, mixed‑type design spaces, but they suffer from low sampling efficiency and often require thousands of evaluations to obtain high‑quality Pareto fronts. To overcome this limitation, the authors propose a hybrid framework that runs BBO in parallel with a large language model (LLM)‑driven sampler.

The design space consists of a serial robot arm with D joints. Each joint can be a Roll, Pitch, or Yaw type, its angle is bounded within