Multi-Innovation Gradient Iterative Locally Weighted Learning Identification for A Nonlinear Ship Maneuvering System

Ship Maneuvering Prediction Using Grey Box Framework via Adaptive RM-SVM with Minor Rudder

“…Therefore, the test data with minor rudder angle was used to identify the RM-SVM. As presented in Table 3, a relatively small amount of training data was used compared to other studies [15] [18] [31], and the validation data was different from the training data. The 20°/20° zigzag test is presented as an example in Figure 4.…”

“…On the basis of multiple tests, Bai used the locally weighted learning to predict maneuvering motion by using the black box structure. However, the maneuver scheme for identification was not suitable as good seamanship handling required [18]. Zhu used a least squared SVM and optimized it with the artificial bee colony method, but the surge velocity was not verified for a zigzag test [19].…”

Full-Scale Maneuvering Trials Correction and Motion Modelling Based on Actual Sea and Weather Conditions

“…Dating back to 1978, Abkowitz utilized Esso Osaka for sea trials, identified the ship maneuvering mathematical model and verified the feasibility of the identification modeling method [2]. Recently, Zhang et al [3], Bai et al [4] and Kim et al [5] also used full-scale ship data for identification modeling. In the literature [2][3][4], it should be noted that the log has also been installed underwater, on the ship hull, which is prone to suffering from cross flow, in addition to the ship being affected by the drift forces of wind and wave.…”

“…Recently, Zhang et al [3], Bai et al [4] and Kim et al [5] also used full-scale ship data for identification modeling. In the literature [2][3][4], it should be noted that the log has also been installed underwater, on the ship hull, which is prone to suffering from cross flow, in addition to the ship being affected by the drift forces of wind and wave. Kim et al [5] employed the method seen in the literature [6][7][8] to correct the sea trial data and identified the ship maneuvering model, but did not consider the influences of wind and waves.…”

Multi‐innovation gradient estimation algorithms for multivariate equation‐error autoregressive moving average systems based on the filtering technique

“…Applying the multi-innovation identification theory into basic recursive identification algorithms such as the least squares algorithm and stochastic gradient algorithm can improve the parameter estimation accuracy [23][24][25]. In this aspect, Bai et al explored a multi-innovation gradient iterative approach for the ship manoeuvring motion with a full-scale trial [26]. Xu derived a multi-innovation stochastic gradient parameter estimation algorithm for the system response from the discrete measurement data by using the moving window data to improve the accuracy of the stochastic gradient identification method [27].…”