In this paper, the classic ship maneuvering equations are adopted and processed to linearized and non-dimensionalized form; and then a kind of Neural Network Recursive Model (NNRM) is derived and applied to the full-scale barge train tests recently conducted in inland waterways in China. The simulation is conducted using three kinds of NNRM training processes described in Section 4.2, Section 4.3 and Section 4.4; and the results are presented and analyzed. By comparing the simulation results with the full-scale tests results, the proposed NNRM has demonstrated powerful ability to predict the barge train maneuvering motion, even in inland waterways.

　

 In Section 4.5, the NNRM is generalized by extending its application to quite a different situation, and the simulation and prediction results are still satisfactory. In Section 4.6 the proposed method is applied to track the trajectory of the barge train. The result shows that there exists a drift. To realize more precisely tracking and trajectory planning [6] using the proposed NNRM is subject to the future work.

　

 In conclusion, it is expected that with further process of environmental effects, external disturbance and consideration of arbitrary velocity change in x direction, the proposed NNRM may be applied to the following objects:

　

1) Course-keeping or automatic pilot and even berthing of barge trains or ships in inland waterways.

　

2) Online, real-time prediction and tracking of barge trains or ships during maneuvering motion in restricted water.

　

 The work presented in this paper is financially supported by the Ministry of Education of China and the National Natural Science Foundation of China under the Grant No.10272085.

　

[1] Xiangjun Kong, Zaojian Zou and Junmin Mou "Practical Methods for Estimating Ship Hydrodynamic Derivatives" (in Chinese), Submitted to Ship Engineering, 2003

[2] Minh-Duc Le, Duc-Hung Nguyen and Kohei Ohtsu "A New Effective Method for Estimation of Ship Hydrodynamic Coefficients", The 4^th Vietnam Conference on Automation, April, 2000

[3] Hess D., Faller W. "Using Recursive Neural Networks for Blind Predictions of Submarine Maneuvers, Proc. of 24^th Symposium on Naval Hydrodynamics, Fukuoka, Japan, Preprints, Vol.IV, pp50-65 July, 2002

[4] Marco D.B., Healey A.J. and Mcghee R.B. "Autonomous Underwater Vehicles: Hybrid Control of Mission and Motion", Automous Robots 3, 169-186, 1996

[5] Daniel Armando Liut "Neural-Network and Fuzzy-Logic Learning and Control of Linear and Nonlinear Dynamic Systems", Dissertation, Faculty of the Virginia Polytechnic Institute and State University, August, 1999

[6] Healey A. J. "Dynamics and Control for UUVs", Naval Postgraduate School Center for AUV Research, ppt, May, 2002

　

 Xiangjun Kong: M.S. course student in Wuhan University of Technology, China. His research interest includes ship course-keeping, intelligent pilot and maneuvering simulation.

　

 Zaojian Zou: Prof. Dr-Ing., is Head of the Institute of CFD in Ship and Ocean Engineering, Wuhan University of Technology, China. His research interest includes prediction of ship maneuverability, calculation of hydrodynamic forces on maneuvering ships and CFD in ship design.

　

 Junmin Mou: Ph.D. course student in Wuhan University of Technology, China. His previous research experience includes ship maneuvering simulation and the safety of navigation.

　

 Yong Li: Prof., is Dean of Navigation School, Wuhan University of Technology, China. His previous research experience includes the safety of navigation and navigation education.