A New Integrated Health Management for Quadrotors Based on Deep Learning

In this paper, the fault diagnosis and health monitoring techniques based on machine learning are investigated before the research. Then, the dynamic model of the quadrotor is analyzed and the most important relationship between the IMU and motor outputs are given. Afterwards, the high-coupling and nonlinear link between units are mapped into an implicit network APN based on LSTM neural network. Predictions and train are made based on the data set collected from the practical and HITL flight log. Finally, the feasibility of the health management is verified during the fault simulation and the a possible solution is given against the common error.