Embodied Generative Method for Scheduling Disinfection Tasks Using Robot

Embodied artificial intelligence (AI) systems, which combine AI with physical robots, have recently received significant attention. However, dynamic environments harbor much uncertainty and unknown situations, posing challenges for robot task scheduling and motion planning. This paper presents an approach that integrates language models with robots to achieve embodied generative method for scheduling disinfection task. We propose a framework that enables the zero-shot capability of disinfection robots in real-world environments. The experiments demonstrate that, compared to traditional static or dynamic planning methods, the proposed framework effectively performs in scenarios with randomly distributed objects and zero-shot situations, thereby enhancing the system's robustness in practical applications.