Preliminary Study of DSL Code Generation for Robotics with LLMs

Code generation through Large Language Models (LLMs) has made significant progress in recent years. However, when the code generation involves a lesser-known Domain-Specific Language (DSL), a standard tool in software development for cyber-physical systems, LLMs’ performance significantly decreases. We present an exploratory study assessing LLMs’ performance in generating LIrAs (Language for Interactive Agents) code, a DSL for robotic and multi-agent tasks specification. This work is a stepping stone towards improving LLM-generated DSL code through iterative specification repair techniques driven by formal verification results.