船舶軌跡追蹤自航器設計法之比較探討

A COMPARATIVE STUDY OF DIFFERENT SHIP TRACK-KEEPING AUTOPILOT DESIGN METHODS

本文旨在探討數學模式需求型，及非數學模式需求型自航器設計法在船舶軌跡追蹤控制應用之結果比較。本文在自航控制器之設計上選擇傳統的數學模式需求型控制器設計法，使用設計簡單且易於實現之內模式控制（IMC,Internal Model Control）及LQR（Linear Quadratic Regulator）兩種方式，而在非數學模式需求型自航器設計法則採用了模糊理論，以及類神經網路兩種方式進行之。一般而言，船舶航行時自航器之實務應用上均存在有環境外力之干擾現象，但為瞭解控制器本身之性能差易特性，故本研究分別以無干擾及有干擾之電腦模擬探討，再以接近真實船舶操縱特性之操船模擬機測試之，用以瞭解干擾對於自航控制器之影響。第一階段將於個人電腦上使用MATLAB之Simulink工作環境作為測試平台，分別於有、無環境外力干擾條件下，以4000噸級軍艦簡化之非線性船舶動態模式作為受控體，分別進行四種不同控制器之軌跡追蹤控制模擬。第二階段則於真時操船模擬機系統上，結合Labview工作環境進行模擬測試，並藉以驗證第一階段於個人電腦進行數值模擬所得之不同控制器性能結果進行分析比較。實驗時使用視線（Line-of-Sight）導航法作為軌跡追蹤之導航法則，用以計算參考航向角，分別於個人電腦及操船模擬機上達成四個路徑點之軌跡追蹤控制。模擬實驗結果顯示，四種（內模式、LQR、模糊、類神經）軌跡追蹤自航器皆能順利完成路徑點追蹤。就軌跡追蹤偏差量而言，於個人電腦所得之四種控制器性能比較結果，略異於操船模擬機之結果。唯於模擬機之實驗條件較為真實，因此以之為主要結論依據，即於無干擾環境下LQR為最佳，IMC及模糊控制次之，而ANN為最差；但於有干擾環境下則反之，即以ANN為最佳，IMC及模糊控制次之，而LQR為最差。為了進行比較，亦於模擬機進行人工操舵軌跡追蹤測試，其用舵較之四種自航控制器更為頻繁。 This work is on the comparison of model-based and model free ship track-keeping autopilots. Specifically, the InternalModel Control (IMC) and Linear Quadratic Regulator (LQR) methods are studied in the model-based approach andthe Fuzzy Logic Control (FLC) and Artificial Neural Network (ANN) are investigated in the model-free approach. First,4way-point tracking simulations are conducted on a personal computer (PC) under the Matlab-Simulink environment, anda simplified nonlinear model of a 4000 ton frigate is adopted as the plant to be controlled with and without environmentaldisturbances. Secondly, same track-keeping experiment scenarios are carried out in the National Taiwan Ocean University(NTOU) shiphandling simulator (SHS) under the Labview working environment. The Line-of-sight (LOS) guidance methodhas been adopted in the simulations to achieve way point tracking through a sequence of course-changing maneuvers.Successful track-keeping maneuvers are achieved with the four types of autopilots (IMC, LQR, FLC and ANN). However,somewhat inconsistent behaviors between the PC-based and SHS-based experiments are noticed in terms of tracking performance.Considering the NTOU SHS provides more realistic experimental conditions, the track-keeping performancecomparisons are based on the results obtained from the SHS. Specifically, with no disturbance effect the LQR is the best,followed by equally well behaved IMC and FLC, and the ANN is the worst. However, the above order is reversed forexperiments under disturbance effect; namely, the ANN outperforms the equally good IMC and FLC, and the LQR is laston the list. SHS-based human pilot track-keeping experiments indicate that much more frequent rudder moments arerecorded as compared with the four types of autopilots.