Abstract:  The unscented Kalman filter （UKF）of GNSS/SINS integrated navigation system is based on the accurate statistical characteristics of measurement noise.When the statistical characteristics of measurement noise change，the filtering performance of UKF will decrease or even diverge if it is not accurately estimated.To solve the above problems，an adaptive UKF algorithm based on improved Sage-Husa （ISHUKF）for GNSS/SINS integrated navigation system is proposed.Firstly，the simplified UKF model of GNSS/SINS nonlinear integrated navigation system is established.Then，an improved Sage-Husa algorithm is proposed based on the analysis of the filter divergence in the conventional Sage-Husa algorithm to ensure the positive quality of the estimation variance of measurement noise.Finally，the simulation experiment of GNSS/SINS integrated navigation system is carried out.The experimental results show that,compared with the Variational Bayesian algorithm，the ISHUKF has roughly the same estimation accuracy for the measurement noise variance，and the algorithm is simpler.Compared with the standard UKF algorithm，the position accuracy，velocity accuracy and attitude accuracy of the integrated navigation system can be improved respectively by about 33%，35% and 72% during the whole simulation period，which verifies the feasibility and superiority of the ISHUKF algorithm and also provides a simple filtering method for integrated navigation system in complex environment.

Key words: improved Sage-Husa algorithm, adaptive UKF, variational Bayesian estimation algorithm, integrated navigation system, measurement noise variance estimation