Identification of the ship maneuvering response model based on recursive refined instrumental variable least-squares

Abstract

To solve the problem of biased estimates of ship maneuvering response model parameters caused by colored noise, this paper presents a recursive refined instrumental variable least-squares algorithm to identify the parameters of the ship maneuvering second-order response model. Based on the recursive instrumental variable least-squares algorithm, the proposed algorithm introduces a dynamic filter to improve the statistical characteristics of colored noise, making the estimated parameter value closely approach the true value of the parameter with colored noise. To verify the effectiveness of the proposed algorithm, performance comparison experiments were performed on the recursive refined instrumental variable least-squares method (RRIVLS) algorithm, recursive instrumental variable least-squares method algorithm, recursive extended least-squares method algorithm, and recursive least-squares method algorithm based on the zigzag test data on a real ship. The results show the RRIVLS algorithm has a better fitting effect and generalization ability on the heading angle than the other three identification models. The corresponding fitting root mean square error and generalization root mean square error are below 1° and 2°, respectively, and the maximum absolute error is below 3° and 4° separately.