Method and system for automatic generation of robot skill primitive library based on large language model

Abstract

The application discloses a robot skill primitive library automatic generation method and system based on a large language model, and relates to the fields of artificial intelligence and robot software engineering. The method comprises the following steps: adopting a double-level function extraction strategy to extract modular functions and process scripts in a robot code library from a robot task-level python file; based on the extracted robot function code, adopting a preset thinking chain prompt word to guide a large language model to generate global semantic description and to distinguish functions with the same name; based on the function metadata of the robot task level after the same name function is distinguished, adopting a preset thinking chain prompt word to guide the configured large language model to perform perception type code reconstruction with embedded field constraints; based on the reconstructed code, adopting a preset thinking chain prompt word to guide the configured large language model to perform adaptive iterative code evaluation and optimization closed loop to obtain a robot skill primitive library. The application can automatically and reliably generate a robot skill primitive library.

Images