Perancangan Sistem Kendali Auto Trajectory di Perairan Laut Jawa pada Kapal Kargo Berbasis Jaringan Saraf Tiruan dengan Model Reference dalam Kondisi Gangguan Arus Laut

Zukhruf, Afra Fathiranis (2026) Perancangan Sistem Kendali Auto Trajectory di Perairan Laut Jawa pada Kapal Kargo Berbasis Jaringan Saraf Tiruan dengan Model Reference dalam Kondisi Gangguan Arus Laut. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009211006-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Proses pengiriman barang melalui jalur laut menjalankan 90% perdagangan global. Laut Jawa merupakan salah satu rute krusial perdagangan. Padatnya jalur laut didukung meningkatnya penggunaan kapal otonom menambah kompleksitas lalu lintas dan risiko kecelakaan. Oleh karena itu, sistem prediksi jalur kapal diperlukan untuk efisiensi dan keselamatan. Penelitian ini merancang sistem auto trajectory kapal menggunakan metode ANN model reference dengan kendali LQR. Model ANN terbaik yang digunakan adalah model ANN dengan 5 hidden layer serta jumlah node 64, 64, 32, 32, dan 16. Hal ini dibuktikan dengan hasil perhitungan mean error Haversine pada titik-titik koordinat dengan menggunakan model ini sebesar 1,17 km dengan menggunakan sistem kendali ANN dan 1,183 km dengan sistem kendali LQR. Pada kondisi tanpa gangguan, sistem ANN model reference mampu mengikuti lintasan desired dengan rata-rata error Haversine 1,176 km dibandingkan dengan sistem LQR yang memiliki rata-rata error Haversine 1,196 km. Kemudian, pada kondisi gangguan arus laut, sistem ANN model reference mampu mengikuti lintasan desired dengan error Haversine 1,928 km, sedangkan untuk sistem LQR pada gangguan aruss laut memiliki error Haversine 1,971 km. Hasil ini menunjukkan sistem ANN model reference lebih baik dari sistem LQR. Dengan total 500 titik uji yang di-sampling setiap 50 menit sepanjang perjalanan 27,78 jam, rata-rata error 1,928 km menunjukkan bahwa sistem kendali dapat mempertahankan performa tracking secara konsisten dalam periode waktu panjang. Tanpa adanya aksi kontrol yang responsif, gaya dorong lateral akibat gangguan arus akan mengakibatkan kapal hanyut (drift) menjauhi lintasan referensi, di mana error akan terakumulasi secara linear seiring bertambahnya waktu dan jarak tempuh. Keberhasilan sistem ANN mempertahankan rata-rata error di angka 1,928 km sepanjang rute Surabaya-Jakarta (~700 km) menunjukkan kontroler aktif memberikan koreksi heading untuk meredam gaya arus dan tidak membiarkan error menjadi tak terkendali seiring waktu.
=================================================================================================================================
Maritime freight transportation accounts for 90% of global trade. The Java Sea is one of the crucial trade routes. The congestion of maritime routes, coupled with the increasing use of autonomous vessels, adds complexity to traffic and increases the risk of accidents. Therefore, a ship trajectory prediction system is necessary for efficiency and safety. This research designs an automatic ship trajectory system using the ANN model reference method with LQR control. The best ANN model employed consists of 5 hidden layers with 64, 64, 32, 32, and 16 nodes respectively. This is evidenced by the mean Haversine error calculation at coordinate points using this model, which yields 1.17 km with the ANN control system and 1.183 km with the LQR control system. Under disturbance-free conditions, the ANN model reference system is able to follow the desired trajectory with an average Haversine error of 1.176 km compared to the LQR system which has an average Haversine error of 1.196 km. Subsequently, under ocean current disturbance conditions, the ANN model reference system is able to follow the desired trajectory with a Haversine error of 1.928 km, whereas the LQR system under ocean current disturbance has a Haversine error of 1.971 km. These results demonstrate that the ANN model reference system outperforms the LQR system. With a total of 500 test points sampled every 50 minutes throughout the 27.78-hour journey, the average error of 1.928 km indicates that the control system can maintain tracking performance consistently over an extended period. Without responsive control action, the lateral thrust force due to current disturbance would cause the ship to drift away from the reference trajectory, where the error would accumulate linearly as time and distance traveled increase. The success of the ANN system in maintaining an average error of 1.928 km along the Surabaya-Jakarta route (~700 km) demonstrates that the controller actively provides heading corrections to mitigate current forces and prevents the error from becoming uncontrolled over time.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ANN, Auto Trajectory, LQR, Sistem Kendali, ANN, Auto Trajectory, Control system, LQR
Subjects:	Q Science > QA Mathematics Q Science > QA Mathematics > QA184 Algebra, Linear Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Afra Fathiranis Zukhruf
Date Deposited:	29 Jan 2026 05:48
Last Modified:	29 Jan 2026 05:48
URI:	http://repository.its.ac.id/id/eprint/131212

Actions (login required)

View Item