Simulasi Pergerakan Pemain Sepakbola Menggunakan Agent-Based Machine Learning

Hidayat, Riyan Martin (2026) Simulasi Pergerakan Pemain Sepakbola Menggunakan Agent-Based Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024201040-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Dalam dunia sepakbola modern, perancangan strategi dan anlisis taktik memerlukan alat bantu visualisasi yang interaktif dan intuitif untuk mengkomunikasikan instruksi permainan kepada para pemain. Pendekatan konvensional dalam pelatihan taktik masih banyak mengandalkan demonstrasi manual di lapangan, yang memakan waktu dan sulit untuk menganalisis sebagai skenario strategi secara sistematis. Penelitian ini mengembangkan simulasi sepakbola dua dimensi (2D) berbasis agen dengan memanfaatkan pendekatan Agent-Based Modeling (ABM) dan Reinforcement Learning menggunakan algoritma Deep Q-Network (DQN). Sistem ini merepresentasikan berbagai peran pemain (penjaga gawang, pemain bertahan, gelandang, penyerang, dan pemain sayap) sebagai agen otonom yang dapat mengambil keputusan taktis secara real-time dalam lingkungan simulasi yang dinamis. Setiap agen dilengkapi dengan jaringan saraf buatan (neural network) yang akan dilatih untuk mempelajari strategi permainan melalui proses pembelajaran berbasis reward. Implementasi menggunakan bahasa pemrorgraman Python dengan pustaka PyTorch untuk deep learning dan Pygame untuk visualisasi grafis 2D. Kerangka kerja yang dikembangkan bersifat modular, sehingga memungkinkan eksperimen dengan berbagai formasi tim (4-3-3, 3-5-2) dan strategi permainan yang berbeda. Penilitian ini diharapkan dapat memberikan kontribusi pada pengembangan alat pelatihan taktik berbasis simulasi dan integrasi teknologi machine learning dalam domain olahraga.
==================================================================================================================================
In modern football, the design of strategies and tactical analysis requires interactive and intuitive visualization tools to communicate game instructions to players. Conventional approaches in tactical training still rely heavily on manual demonstrations on the field, which are time-consuming and difficult for systematic analysis of various strategy scenarios. This research develops a two-dimensional (2D) football simulation based on agents by utilizing Agent-Based Modeling (ABM) and Reinforcement Learning approaches using the Deep Q-Network (DQN) algorithm. The system represents various player role (goalkeeper, defender, midfielder, striker, winger) as autonomous agents capable of making tactical decision in real-time within a dynamic simulation environment. Each aganet is equipped with an artifical neural network trained to learn game strategies through reward-based learning processes. Implementation uses Python programming language with PyTorch libraries for deep learning and Pygame for 2D graphics visualization. The developed framework is modular in nature, allowing experiments with various team formations (4-3-3, 3-5-2, etc) and diferent game strategies. This research is expected to contribute to the development of simulation-based tactical training tools and the integration of machine learning technology in the sports domain.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Agent-Based Modeling, Deep Q-Network, Multi-Agent Reinforcement Learning, Simulasi Sepakbola 2D, Pengambilan Keputusan Taktis, 2D Football Simulation, Tactical Decision-Making
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.62 Simulation
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Riyan Martin Hidayat
Date Deposited:	26 Jan 2026 09:23
Last Modified:	26 Jan 2026 09:23
URI:	http://repository.its.ac.id/id/eprint/130185

Actions (login required)

View Item