Best Practice of Scenario-Based Modeling for Humanoid Software Validation

초록

Recent advances in adapting artificial intelligence technology have revolutionized the field of humanoid robots. The intelligence of robots is based on reinforcement learning and AI models. However, due to the nature of robot software that continuously learns and reacts through interaction with the real environment, it isn’t easy to validate safety, reliability, and accuracy using only existing software testing methodologies. To address these issues, we apply scenario-based modeling in software validation to humanoid robot software, incorporating reinforcement learning models. In particular, the ‘mode selection’ function of humanoid robots is presented as an application case, and a systematic procedure is performed from requirement analysis to scenario creation, refined scenario-based test case creation, and scenario execution and validation using a behavior tree. This is expected to contribute to maintaining the stability of humanoid robot systems by systematically detecting unpredictable behaviors in humanoid software and enhancing efficiency and reliability by validating the decision-making process of AI models using scenario-based modeling based on behavior trees. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

키워드