Abstract: A quaternion particle filter algorithm was proposed to solve the aerial vehicle nonlinear attitude determination problem. To reduce computational burden, the state vector was divided into linear part and nonlinear part. Aiming at quaternion weighting sum normalization problem, a Lagrange cost function was derived to compute the predicted quaternion by minimizing the cost function, and the normalization of disturbed quaternion was ensured by calculating the square root of diagonal elements of the quaternion error covariance matrix, the multiplicative quaternion error was used for predicted covariance computation of the quaternion, since it represented the distance between quaternion points and the predicted mean quaternion. The simulation results show that, compared to traditional extended Kalman filter（EKF） and unscented Kalman filter（UKF）algorithms, the proposed algorithm has higher estimation precision and better stability.