Artificial Neural Network controller for automatic ship berthing: challenges and opportunities

Abstract

Ship berthing is a complicated manoeuvring process that demands precise control of the speed and course of an underactuated system. As berthing involves low speed running of a ship under environmental disturbances with reduced manoeuvrability, a professional ship handler often faces difficulties in controlling heading in such a situation. To bring automation in ship berthing, most of the researchers have agreed that Artificial Neural Network (ANN) plays a vital role as it has the ability to learn from human experience and replicate similar action in an unknown situation. However, we are still far away from implementing it for real berthing control as we do not have any self-fulfilling ANN controller yet, which can treat all the major issues relevant to a complicated ship berthing operation. Based on contemporary research findings, this paper, therefore, highlights four major challenges that have to be taken into account while proposing an ANN controller for ship berthing, and a comprehensive summary of how to deal with those. The first is how to provide consistent teaching data while training ANN controller to make it more robust; second is how to make the controller universal to do berthing in any port; third is how to tackle the wind disturbances while automation in progress; and the fourth is how to align a ship to the pier, which is the final stage of berthing.

State-of-the-Art Review and Future Perspectives on Maneuvering Modeling for Automatic Ship Berthing

“…Automatic berthing control studies [62][63][64][65][66] considering the berthing maneuver characteristics have shown satisfactory results in comparison of model tests. In these research studies, the berthing maneuver specifications are discussed: Hydrodynamic forces acting on the ship have strong non-linearity; the ship longitudinal velocity is small, and is of the same order as the lateral velocity and yaw moment.…”

Automatic berthing of unmanned surface vessels with predetermined performance

“…Further, since 𝑙𝑖𝑚 𝑡→∞ 𝑒 −𝜛𝑡 = 0 . 𝑍 converges to compact set 𝛺 = {𝑧 ∈ 𝑅 3 , ‖𝑧‖ ≤ √2𝜇/𝜛}, by selecting the design parameters of 𝛽 1 , 𝛽 2 , 𝛬 𝑐 , 𝛤, 𝜅, 𝜒 to arbitrarily small to ensure formula (36) -( 38) is founded. In addition, according to formula (12), it can be found that |𝑒 1,𝑖 | < 𝜎 𝑖 is valid for ∀𝑡 > 0, that is, formula ( 11) is valid.…”

“…In addition to common wind, wave and current interference, vessels are also affected by shore wall effect and shallow water effect [1] during automatic berthing, which makes automatic berthing control more difficult than common tracking control. Scholars have adopted a variety of methods, including backstep method [2] and neural network [3], to study automatic berthing of vessels, and achieved good results.…”

“…𝜉 1 (𝑧)𝑇 , 𝜉 1 (𝑧)𝑇 , 𝜉 1 (𝑧) 𝑇 ] 𝑇 . 𝜀 𝑐 is the approximate error vector of the neural network, and 𝜀 𝑐 ∈ 𝑅3 satisfies ‖𝜀 𝑐 ‖ ≤ 𝜀̄𝑐, where 𝜀̄𝑐is a constant.Let 𝑑 𝑐 = 𝜀 𝑐 + 𝜏 𝑑 , and then let assuming 1 and lemma 3 have an unknown vector 𝜃 = [𝜃 1 , 𝜃 2 , 𝜃 3 ] 𝑇 leads to 𝜃 𝑖 > |𝑑 𝑐,𝑖 |, (𝑖 = 1,2,3),𝜃 𝑖 is a normal number. According to lemma 1, we can get:…”

Automatic berthing of unmanned surface vessels with predetermined performance

“…In addition to common wind, wave and current interference, vessels are also affected by shore wall effect and shallow water effect 1 during automatic berthing, which makes automatic berthing control more difficult than common tracking control. Domestic scholars have adopted a variety of methods, including backstep method 2 and neural network 3 , to study automatic berthing of vessels, and achieved good results.…”