Application of unscented Kalman filter (UKF) algorithm in solving two-point boundary value problem (TPBVP) is a key technology in the indirect method of optimization problem solving algorithms,and usually the final solution can only be given through numerical iteration methods.Due to the high sensitivity and strong nonlinear dynamic characteristics of the two-point boundary value problem,the accuracy of the initial value of the numerical iteration is required to be very high,otherwise it is difficult to obtain a convergent solution.In order to reduce the requirements of the initial value accuracy,based on the UKF algorithm,an optimal parameter estimation method is given,which can be used to solve the two-point boundary value problem.This method adopts Gaussian distribution model,which can describe the original nonlinear problem with accuracy of second order or higher,and has a larger convergence range and stronger robustness.The example of spacecraft trajectory optimization verifies the effectiveness and reliability of this algorithm.