Abstract: Target tracking in dynamic positioning observation equations is nonlinear and stochastic model with uncertainty, and goals in the process of movement by the random disturbance is larger. It is difficult to determine a priori variance, which may lead to the update parameter estimation errors in iterative process, which can lead to filter divergence. According to the above problem, this paper presents an improved adaptive square root unscented particle filter algorithm(ASRUPF). This algorithm combines the adaptive filtering theory, square root unscented kalman filter (SRUKF) and particle filter (PF), which determine the system measurement and the probability density function of state noise, to ensure variance matrix is non-negative qualitative. The algorithm effectively improves the single point dynamic positioning accuracy.

Key words: SRUPF, adaptive, nonlinear kalman, dynamic positioning