WIBOWO, BETRIK SESYANTO HADI (2026) PROTOTIPE UNMANNED SURFACE VEHICLE BERBASIS PERCEPTION-DRIVEN AUTONOMY DENGAN REINFORCEMENT LEARNING LITE (Q-TABLE) UNTUK LAKE SURFACE DEBRIS RECOVERY. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
Cover.pdf Download (3MB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
Bab 1.pdf Restricted to Registered users only Download (317kB) |
|
|
Text (BAB II)
Bab 2.pdf Restricted to Registered users only Download (759kB) |
|
|
Text (BAB III)
Bab 3.pdf Restricted to Registered users only Download (1MB) |
|
|
Text (BAB IV)
Bab 4.pdf Restricted to Registered users only Download (1MB) |
|
|
Text (BAB V)
Bab 5.pdf Restricted to Registered users only Download (170kB) |
|
|
Text (DAFTAR PUSTAKA)
Daftar Pustaka.pdf Restricted to Registered users only Download (193kB) |
|
|
Text (LAMPIRAN)
Lampiran.pdf Restricted to Registered users only Download (1MB) |
Abstract
Plastic waste in Indonesian inland waters requires an efficient and sustainable cleanup solution. This study designs, builds, and tests an Unmanned Surface Vehicle (USV) prototype based on Perception-Driven Autonomy to autonomously detect and collect plastic waste, integrating an adaptive perception system, Reinforcement Learning Lite (Q-Table) navigation, and hybrid solar energy support. The wave-piercing pontoon structure, built from EPS density 20 foam layered with resin-fiberglass, a C75 lightweight steel frame, and a 75×50×30 cm gridwall basket, produced a rigid construction with no significant dimensional distortion. HuskyLens camera calibration across 27 test scenarios (1,350 frames, 14,850 data points) recorded an average Detected score of 0.98, Loop Time of 1.1 ms, and Stable score of 0.94, with perfect detection on colored plastic bottles. The TF-Luna LiDAR sensor, tested across 18 scenarios (900 frames, 2,700 data points), showed high accuracy and immunity to lighting variation between 7 and 495 lux, with a maximum STDEV of 3.16 cm at a 15° incidence angle. The u-blox NEO-M8N GPS module was validated against a Topcon GTS-235N Total Station, achieving 0.48-1.29 m positional accuracy and 0.69-1.84 HDOP, substantially outperforming an iPhone 14 Pro Max (13.24 m). A 100 WP solar panel with a 30A MPPT controller reached a 97.85% conversion efficiency, supporting a Home Base Return feature that navigated the vehicle back to its starting point within a tolerance radius of under 7 m. Q-Learning training over 12,000 episodes produced an optimal policy with a 98% success rate and an average of 26.6 steps per episode, achieving 100% validated navigation success across five randomized 20×20 grid maps. Field testing on a lake confirmed an effective basket capacity of 97,500 cm³, equivalent to approximately 114 plastic bottles, with a passive collection mechanism operating without additional actuators. These results demonstrate that the USV prototype can operate functionally, autonomously, and sustainably in lake surface debris recovery missions. Keywords: Unmanned Surface Vehicle, Perception-Driven Autonomy, Reinforcement Learning Lite (Q-Table), lake surface debris recovery Sampah plastik di perairan darat Indonesia memerlukan solusi pembersihan yang efisien dan berkelanjutan. Penelitian ini merancang, membangun, dan menguji prototipe Unmanned Surface Vehicle (USV) berbasis Perception-Driven Autonomy untuk mendeteksi dan mengumpulkan sampah plastik secara otonom, mengintegrasikan sistem persepsi adaptif, navigasi Reinforcement Learning Lite (Q-Table), dan energi hybrid surya. Struktur ponton wave-piercing dari EPS density 20 berlapis resin-fiberglass, rangka baja ringan profil C75, dan keranjang gridwall 75×50×30 cm menghasilkan konstruksi kokoh tanpa distorsi dimensi berarti. Kalibrasi kamera HuskyLens pada 27 skenario (1.350 frame, 14.850 data) mencatat rata-rata Detected 0,98, Loop Time 1,1 ms, dan Stable 0,94, dengan deteksi sempurna pada botol plastik berwarna. Sensor TF-Luna LiDAR pada 18 skenario (900 frame, 2.700 data) menunjukkan akurasi tinggi dan ketahanan terhadap variasi pencahayaan 7-495 lux, dengan STDEV maksimum 3,16 cm pada sudut insidensi 15°. Modul GPS u-blox NEO-M8N tervalidasi terhadap Total Station Topcon GTS-235N dengan akurasi 0,48-1,29 m, HDOP 0,69-1,84, jauh melampaui iPhone 14 Pro Max (13,24 m). Panel surya 100 WP dengan SCC MPPT 30A mencapai efisiensi konversi 97,85%, mendukung fitur Home Base Return yang menavigasi wahana kembali ke titik awal dalam radius toleransi kurang dari 7 m. Pelatihan Q-Learning selama 12.000 episode menghasilkan kebijakan optimal dengan success rate 98% dan rata-rata 26,6 langkah per episode, tervalidasi 100% pada lima peta grid 20×20 acak. Pengujian lapangan di perairan danau membuktikan kapasitas efektif keranjang 97.500 cm³, setara ±114 botol plastik, dengan mekanisme pengumpulan pasif yang bekerja tanpa aktuator tambahan. Hasil ini membuktikan prototipe USV mampu beroperasi secara fungsional, otonom, dan berkelanjutan dalam misi lake surface debris recovery. Kata Kunci: Unmanned Surface Vehicle, Perception-Driven Autonomy, Reinforcement Learning Lite (Q-Table), lake surface debris recovery
| Item Type: | Thesis (S1) |
|---|---|
| NIM/NIDN Creators: | 41421010016 |
| Uncontrolled Keywords: | Unmanned Surface Vehicle, Perception-Driven Autonomy, Reinforcement Learning Lite (Q-Table), lake surface debris recovery |
| Subjects: | 600 Technology/Teknologi > 620 Engineering and Applied Operations/Ilmu Teknik dan operasi Terapan > 621 Applied Physics/Fisika terapan > 621.3 Electrical Engineering, Lighting, Superconductivity, Magnetic Engineering, Applied Optics, Paraphotic Technology, Electronics Communications Engineering, Computers/Teknik Elektro, Pencahayaan, Superkonduktivitas, Teknik Magnetik, Optik Terapan, Tekn 700 Arts/Seni, Seni Rupa, Kesenian > 740 Drawing and Decorative Art/Menggambar dan Seni Dekorasi > 748 Glass/Kaca > 748.8 Specific Glass Arts/Seni Kaca Khusus > 748.82 Bottes/Botol |
| Divisions: | Fakultas Teknik > Teknik Elektro |
| Depositing User: | khalimah |
| Date Deposited: | 22 Aug 2026 02:14 |
| Last Modified: | 22 Aug 2026 02:14 |
| URI: | http://repository.mercubuana.ac.id/id/eprint/103343 |
Actions (login required)
 |
View Item |
