Aiming at precision problem of current unmanned aerial vehicle（UAV）attitude solution

a dynamic step size gradient descent attitude solution algorithm based on Root Mean Square Propagation（RMSprop） algorithm is proposed.The improved algorithm firstly uses the gyroscope to solve output data directly from quaternion differential equation

measures the three-axis Euler angle change rate in object coordinate system to obtain attitude quaternion rate

and then uses the RMSprop algorithm to calculate the attitude.Based on the above theory

multi-sensor data fusion is adopted

and dynamic step optimizing estimation of the attitude quaternion is performed through the RMSprop algorithm to enhance dynamic performance of attitude solution of the small UAV aircraft

and improve flight performance and anti-interference performance and reliability.The experimental results show that the algorithm has higher attitude solution precision and better real-time tracking performance.