Optimasi Rute Pelayaran Kapal Menggunakan Reinforcement Learning Dan Mix Interger Linear Programming

Risqi, Mohammad Daffa Athalla (2025) Optimasi Rute Pelayaran Kapal Menggunakan Reinforcement Learning Dan Mix Interger Linear Programming. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5018211018-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Dekarbonisasi merupakan gerakan global yang penting untuk mengurangi emisi karbon dengan beralih ke sumber energi terbarukan dan berkelanjutan serta mengurangi penggunaan bahan bakar fosil. Dalam industri perkapalan, yang mengangkut lebih dari 90% perdagangan global, emisi karbon telah menjadi masalah yang besar, dengan proyeksi menunjukkan bahwa emisi dari sektor perkapalan akan meningkat hingga 130% pada tahun 2050. Untuk mengatasi tantangan ini, sangat penting untuk fokus pada strategi optimasi rute kapal guna meminimalkan konsumsi bahan bakar, yang merupakan kontributor utama terhadap emisi karbon. Kebutuhan untuk optimasi rute kapal ini sangat mendesak untuk mencapai tujuan dekarbonisasi yang tercantum dalam perjanjian internasional seperti Initial IMO GHG Strategy 2018, yang menekankan pengurangan emisi gas rumah kaca dari industri perkapalan. Penelitian ini mengusulkan model berbasis simulasi untuk kapal kontainer, yang mengintegrasikan Reinforcement Learning (RL), khususnya Deep-Q Network (DQN) sebagai algoritma, dengan Mixed Integer Linear Programming (MILP) untuk mencapai rute optimal yang meminimalkan konsumsi bahan bakar dan waktu tempuh, sambil mempertimbangkan kendala dinamis waktu nyata seperti kondisi cuaca, arus laut, dan hambatan navigasi. Sistem ini memanfaatkan DQN untuk memproses input dinamis seperti karakteristik kapal, data cuaca, dan kondisi lingkungan untuk memprediksi konsumsi bahan bakar saat rute berkembang dan menghasilkan rute optimal yang dapat secara signifikan mengurangi emisi karbon dan meningkatkan efisiensi bahan bakar, menjadikan hal yang solutif dan menjanjikan untuk mendukung program dekarbonisasi danmembantu kru kapal dalam menentukan rute optimal ketika berlayar.
==================================================================================================================================
Decarbonization represents a critical global initiative that aimed at reducing carbon emissions by transforming to renewable and sustainable energy sources while phasing out the use of fossil fuels. In the maritime industry, where over 90% of global trade is carried by ships, carbon emissions have been a growing concern, with projections indicating that emissions from the shipping sector will increase by up to 130% by 2050. To address this challenge, it is essential to focus on optimization strategies for ship route in order to minimize fuel consumption, a major contributor to carbon emissions. The need for ship route optimization is particularly urgent to achieve the decarbonization goals outlined in global agreements such as the Initial IMO GHG Strategy 2018, which emphasizes reducing greenhouse effect as a gas emissions from the maritime industry. This study proposes a simulation-based model for container ship, integrating a Reinforcement Learning (RL), particularly Deep-Q Network (DQN) for the algorithm, with Mixed Integer Linear Programming (MILP) to achieve an optimal route that minimizes fuel consumption and travel time while addressing real-time dynamic constraints such as weather conditions, sea currents, and navigational obstacles. DQN, as a reinforcement learning algorithm, is chosen for its ability by handling a discrete decisions, such as selecting waypoints along a ship route is based on the current state of the environment. The integration of MILP ensures that the optimization process is computationally efficient and provides a robust solution that aligns with the real-world constraints of the shipping industry. The system utilizes DQN to process dynamic inputs such as ship characteristics, weather data, and environmental conditions to predict fuel consumption as the route evolves, ensuring that the model can adapt to sudden changes in circumstances. Through iterative training, the system generates optimal shipping routes that can significantly reduce carbon emissions and improve fuel efficiency, making it a promising solution for supporting maritime decarbonization and contributing to help the crew of ship when decide an optimal route

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Optimasi, Reinforcement Learning, Deep-Q Network, Mixed Integer Linear Programming, Rute Pelayaran, Optimization, Reinforcement Learning, Deep-Q Network, Mixed Integer Linear Programming, Ship Route.
Subjects:	Q Science > QA Mathematics > QA9.58 Algorithms V Naval Science > VK > VK570 Optimum ship routing. V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM276.A1 Fuel (Including supplies, costs, etc.)
Divisions:	Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User:	Mohammad Daffa Athalla Risqi
Date Deposited:	05 Aug 2025 09:56
Last Modified:	05 Aug 2025 09:56
URI:	http://repository.its.ac.id/id/eprint/127175

Actions (login required)

View Item