Multimodal modeling and identification study of a quadrotor robot

Abstract The flight control system of a flying robot plays a crucial role. Therefore, it is imperative to establish various flight mode control models. Traditional modeling methods face challenges in meeting the demands of rapid research and development due to the emergence of diverse aircraft shapes and load forms. To address these issues, this study focuses on multi-conditions modeling and system identification. Based on flight dynamics theory, a comprehensive quadrotor robot dynamics model along with its state space model was established. Flight tests were designed for system identification under different flight modes. The data obtained from the flight test was preprocessed using an extended Kalman filter algorithm based on flight path reconstruction technology to reduce constant deviation from airborne sensors. Frequency response identification technology was employed to identify control models for hovering mode and forward flight at 10m/s mode in the quadrotor robot. Model structures were determined based on precision index evaluation criteria. Time-domain validation method was utilized to simulate fitting error RMS for each channel model under multiple conditions using different signal excitations derived from datasets collected during experiments. The results confirmed that the proposed aircraft system identification method is effective and feasible.