Ramadhan, Muhammad Gemilang (2026) Implementasi Deep Deterministic Policy Gradient Sebagai Parameter Tuner Lazy Theta* dan Safe Artificial Potential Field untuk Perencanaan Lintasan pada USV LSS-01. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5022211184-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5022211184-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy |
Abstract
Perencanaan lintasan merupakan elemen penting pada Unmanned Surface Vehicle (USV) untuk menghasilkan navigasi yang aman, efisien, dan adaptif. Penelitian ini mengembangkan sistem perencanaan lintasan USV LSS-01 menggunakan Lazy Theta* sebagai global path planner dan Safe Artificial Potential Field sebagai local path planner. Deep Deterministic Policy Gradient diterapkan sebagai adaptive parameter tuner untuk mengatur bobot heuristic dan bobot clearance pada Lazy Theta*, serta gain attractive, gain repulsive, dan jarak aman pada Safe Artificial Potential Field. Pelatihan dan pengujian dilakukan menggunakan simulasi model USV empat derajat kebebasan, kemudian diimplementasikan pada USV LSS-01 di Kolam InfinITS ITS. Hasil simulasi Lazy Theta*-DDPG menghasilkan jarak lintasan 50,3 m dan waktu tempuh 37,8 s, lebih baik daripada parameter tetap terbaik dengan jarak 50,6 m dan waktu 38,0 s. Pada pengujian dengan rintangan dinamis, SAPF-DDPG menghasilkan jarak lintasan 51,3 m dan waktu tempuh 45,0 s. Hasil implementasi menghasilkan jarak lintasan 36,94 m tanpa rintangan dinamis dan 39,89 m dengan rintangan dinamis. Hasil tersebut menunjukkan bahwa DDPG mampu menyesuaikan parameter kedua algoritma secara adaptif sehingga USV dapat mencapai waypoint dan menghindari rintangan statis maupun dinamis.
======================================================================================================================================
Path planning is an important element of an Unmanned Surface Vehicle (USV) to ensure safe, efficient, and adaptive navigation. This study develops a path-planning system for the LSS-01 USV using Lazy Theta* as the global path planner and Safe Artificial Potential Field as the local path planner. Deep Deterministic Policy Gradient (DDPG) is applied as an adaptive parameter tuner for both algorithms. In Lazy Theta*, DDPG adjusts the heuristic and clearance weights, while in the Safe Artificial Potential Field, it adjusts the attractive gain, repulsive gain, and safe distance according to the navigation conditions. Training and testing were conducted using a four-degree-of-freedom USV simulation model, followed by implementation on the LSS-01 USV at the InfinITS ITS Pond. The Lazy Theta*-DDPG simulation produced a path length of 50.3 m and a travel time of 37.8 s, outperforming the best fixed-parameter configuration, which produced a path length of 50.6 m and a travel time of 38.0 s. In testing with dynamic obstacles, SAPF-DDPG produced a path length of 51.3 m and a travel time of 45.0 s. The implementation results produced path lengths of 36.94 m without dynamic obstacles and 39.89 m with dynamic obstacles. These results demonstrate that DDPG can adaptively adjust the parameters of both algorithms, enabling the USV to reach the designated waypoints while avoiding static and dynamic obstacles.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | DDPG, Lazy Theta Star, Perencanaan Lintasan, SAPF, USV |
| Subjects: | Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA9.64 Fuzzy logic T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL152.8 Vehicles, Remotely piloted. Autonomous vehicles. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis |
| Depositing User: | Muhammad Gemilang Ramadhan |
| Date Deposited: | 28 Jul 2026 02:31 |
| Last Modified: | 28 Jul 2026 02:31 |
| URI: | http://repository.its.ac.id/id/eprint/137491 |
Actions (login required)
 |
View Item |