A Self-Evolving Digital Twin System Architecture for Indoor UAV Management Based on Online Learning

In recent years, there has been a growing interest in the concept of Digital Twin (DT), which serves as a powerful tool for digital modeling the physical world and feedback to reality. The key problem of digital twin modeling in robotics is how to represent the physical entities accurately and consis-tently in a dynamic environment, as well as give timely guidance to physical entities by DT model. Therefore, we propose a 4D (four-dimension) digital twin system framework for mission generation, energy prediction, and predictive decision-making during the autonomous indoor inspection mission in this work. Further, to cope with the model drift problem in the energy prediction DT model, an online learning method is used to adapt to the dynamic flight conditions during actual flight. The efficacy of the proposed method is validated through actual flight experiments.