Pengaturan Pergerakan Kerumunan Menggunakan Reinforcement Learning dalam Simulasi Evakuasi Kapal: Studi Kasus pada Perilaku NPC Jamak

Prakosa, Ilham Jalu (2025) Pengaturan Pergerakan Kerumunan Menggunakan Reinforcement Learning dalam Simulasi Evakuasi Kapal: Studi Kasus pada Perilaku NPC Jamak. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022222008-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2027. Download (2MB) | Request a copy

Abstract

Dengan semakin meningkatnya kompleksitas simulasi kapal, diperlukan pendekatan kecerdasan buatan yang canggih untuk meningkatkan efisiensi dan keselamatan proses evakuasi. Proposal tesis ini bertujuan untuk mengembangkan sistem kecerdasan buatan berbasis Reinforcement Learning (RL) untuk mengelola evakuasi dalam lingkungan simulasi kapal. Fokus utama adalah pada perilaku Non-Player Characters (NPC), yang berperan sebagai agen virtual yang harus diarahkan ke tempat evakuasi. Metodologi penelitian ini melibatkan beberapa tahap. Pertama, dilakukan pembuatan lingkungan simulasi kapal yang realistis untuk menyajikan suatu konteks yang mendekati kondisi nyata. Selanjutnya, model Reinforcement Learning (RL) dibangun, dan dilakukan pelatihan model. Pada tahap pelatihan, dua metode yang akan dibandingkan, yaitu Proximal Policy Optimization (PPO) dan Soft ActorCritic (SAC), akan digunakan untuk memperoleh kebijakan optimal dalam evakuasi. Setelah tahap pelatihan, implementasi sistem dilakukan diikuti oleh evaluasi performa dalam konteks studi kasus NPC. Diharapkan bahwa hasil dari penelitian ini tidak hanya akan memberikan pemahaman yang lebih mendalam tentang kecerdasan buatan dalam konteks evakuasi kapal, tetapi juga akan memiliki aplikasi praktis yang signifikan, terutama pada simulasi perilaku NPC dalam situasi darurat kapal. Kombinasi teknologi RL dengan simulasi kapal membuka peluang baru untuk meningkatkan pengalaman dan efisiensi pelatihan evakuasi, serta memperkuat kemampuan kapal dalam menghadapi situasi darurat. Proposal ini memadukan teknologi tinggi untuk mencapai tujuan evakuasi otomatis yang efektif, dengan harapan dapat meningkatkan keselamatan dan respons dalam dunia pelayaran.
================================================================================================================================
As the complexity of ship simulations continues to grow, sophisticated artificial intelligence approaches are required to enhance the efficiency and safety of evacuation processes. This thesis proposal aims to develop a Reinforcement Learning (RL)-based artificial intelligence system to manage evacuation in ship simulation environments. The primary focus lies on the behavior of Non-Player Characters (NPCs), serving as virtual agents directed towards evacuation points. The research methodology comprises several crucial stages. Initially, a realistic ship simulation environment is created to provide a context closely resembling real-world conditions. Subsequently, a Reinforcement Learning (RL) model is constructed, and the model undergoes training. During the training phase, two methods—Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC)—will be employed and compared to obtain optimal evacuation policies. Following the training phase, system implementation is conducted, follow d by performance evaluation within the context of NPC case studies. It is anticipated that the outcomes of this research will not only deepen the understanding of artificial intelligence in the context of ship evacuation but will also have significant practical applications, particularly in simulating NPC behavior during ship emergency situations. The integration of RL technology with ship simulation opens new opportunities to enhance the training efficiency and experience of evacuations while reinforcing a vessel’s capabilities in responding to emergency situations. This proposal combines advanced technology to achieve effective automated evacuation goals, with the hope of improving safety and responsiveness in the maritime industry.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	NPC, Proximal Policy Optimization, Reinforcement Learning, Soft Actor-Critic
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) T Technology > T Technology (General) > T55 Industrial Safety T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Ilham Jalu Prakosa
Date Deposited:	23 Jan 2025 07:12
Last Modified:	23 Jan 2025 07:12
URI:	http://repository.its.ac.id/id/eprint/116731

Actions (login required)

View Item