In order to solve the problem of inconsistent state estimation of EKF-SLAM algorithm, an improved observability EKF algorithm was proposed. The first available estimate of each state variable was used to calculate the Jacobian matrix of the filter, thus the observable subspace of the error state system model was guaranteed to have the same dimension as the observable subspace of the actual nonlinear SLAM system. In the case of unknown measurement noise covariance, the initial exponential moving average method was used to estimate the covariance of the measurement noise for the EKF-SLAM. The results show that the improved EKF-SLAM algorithm is more accurate than the traditional EKF-SLAM algorithm, and the mean square error is reduced by at least 30%.