Pengembangan Algoritma Optimasi Rute Unmanned Aerial Vehicle Menggunakan Algoritma Genetika – Variable Neighborhood Search Berbasis Reinforcement Learning Pada Ladang Pertanian

Cristianto, Nikolaus Vico (2026) Pengembangan Algoritma Optimasi Rute Unmanned Aerial Vehicle Menggunakan Algoritma Genetika – Variable Neighborhood Search Berbasis Reinforcement Learning Pada Ladang Pertanian. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6026251024-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (15MB) | Request a copy

Abstract

Pertanian presisi menuntut perencanaan operasional yang efisien untuk meningkatkan produktivitas serta mengurangi pemborosan sumber daya, khususnya dalam penggunaan Unmanned Aerial Vehicle (UAV) untuk penyemprotan atau penyiraman tanaman. Permasalahan optimasi rute UAV memiliki tingkat kompleksitas tinggi karena dipengaruhi oleh keterbatasan kapasitas UAV, bentuk lahan yang tidak beraturan, serta kebutuhan penyemprotan pada area kritis (stressed regions); sehingga permasalahan ini dapat dimodelkan sebagai Capacitated Vehicle Routing Problem (CVRP). Penelitian ini telah mengembangkan pendekatan optimasi rute UAV yang adaptif dan stabil melalui algoritma hibrida Genetic Algorithm (GA) dan Variable Neighborhood Search (VNS), yang diperkuat dengan mekanisme Reinforcement Learning berbasis Upper Confidence Bound (UCB) sebagai pengendali adaptif dalam pemilihan operator pencarian lokal (neighborhood). Evaluasi telah dilakukan melalui simulasi menggunakan dataset standar dan skenario uji terkontrol. Hasil simulasi menunjukkan bahwa algoritma Hibrida GA-VNS-UCB yang diusulkan mengungguli metode metaheuristik konvensional dalam hal kualitas solusi, efisiensi komputasi, dan stabilitas pencarian. Lebih lanjut, kemampuan belajar dinamis dari agen UCB secara efektif berhasil mencegah konvergensi prematur (jebakan local optima) dengan menyeimbangkan eksplorasi dan eksploitasi operator pencarian lokal secara optimal. Pada akhirnya, pendekatan ini memberikan kontribusi metodologis yang kuat serta solusi berskalabilitas tinggi untuk pengerahan UAV otonom dalam operasi pertanian presisi.
============================================================
============================================================
==========
Precision agriculture demands efficient operational planning to increase productivity and reduce resource waste, particularly in the use of Unmanned Aerial Vehicles (UAVs) for crop spraying or watering. UAV route optimization problems possess a high level of complexity as they are influenced by UAV capacity constraints, irregular field shapes, and spraying requirements in stressed regions; thus, they can be modeled as a Capacitated Vehicle Routing Problem (CVRP). This study developed an adaptive and stable UAV route optimization approach through a hybrid Genetic Algorithm (GA) and Variable Neighborhood Search (VNS), reinforced with a Reinforcement Learning mechanism based on the Upper Confidence Bound (UCB) as an adaptive controller for neighborhood operator selection. The evaluation was conducted through simulations using standard datasets and controlled test scenarios. The simulation results demonstrate that the proposed Hybrid GA-VNS-UCB algorithm outperforms conventional metaheuristic methods in terms of solution quality, computational efficiency, and search stability. Furthermore, the dynamic learning capability of the UCB agent effectively mitigates premature convergence by optimally balancing the exploration and exploitation of local search operators. Ultimately, this approach provides a robust methodological contribution and a highly scalable solution for autonomous UAV deployment in precision agriculture.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	pertanian presisi, UAV, CVRP, algoritma genetika, variable neighborhood search, upper confidence bound, reinforcement learning ============================================================ precision agriculture, UAV, CVRP, genetic algorithm, variable neighborhood search, upper confidence bound, reinforcement learning
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions:	Faculty of Information and Communication Technology > Information Systems > 59101-(S2) Master Thesis
Depositing User:	Nikolaus Vico Cristianto
Date Deposited:	24 Jul 2026 02:40
Last Modified:	24 Jul 2026 02:40
URI:	http://repository.its.ac.id/id/eprint/136718

Actions (login required)

View Item