Hidayatullah, Muhammad Lukmanul Hakim Hidayatullah (2026) Berikut adalah judul tersebut dengan kapital di awal setiap kata (Title Case): Perancangan Sistem Hibrida Neural Network-Genetic Algorithm untuk Penyelesaian Kinematika Balik Posisi End-Effector Robot Open Manipulator-X. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5022221055-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5022221055-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy |
Abstract
Penelitian ini mengusulkan sistem Hibrida Neural Network–Genetic Algorithm (NN-GA) untuk menyelesaikan permasalahan kinematika balik secara akurat dengan kasus yang digunakan adalah robot OpenManipulator-X 4-DoF yang bersifat redundan terhadap tugas posisi. Sistem hibrida yang diusulkan mengintegrasikan kecepatan komputasi Neural Network (NN) dan keakuratan Genetic Algorithm (GA). Pada Tahap I , terdapat proses GA Tuner untuk mengoptimasi hyperparameter tiga model NN spesialis secara otomatis. Pada Tahap II, prediksi paralel ketiga NN diseleksi melalui mekanisme Anchor Selection (pemilihan solusi NN terbaik) berbasis evaluasi Forward kinematics (FK), kemudian diperhalus oleh GA Refinement hingga mencapai akurasi target sub-milimeter. Sistem divalidasi pada tiga level bertingkat. Pada Level 1, yakni pengujian murni hasil prediksi sistem hibrida, sistem berhasil mencapai Success Rate (SR) 100% pada 5 titik tunggal (mean error 0,0891 mm) dan SR 95,80% pada 500 titik acak. Studi komparasi 20 titik menunjukkan Hibrida NN-GA menghasilkan mean error 0,0769 mm dan SR 100%, lebih unggul dibandingkan NN-Only (mean error 59,54 mm) dan Simple GA (mean error 0,67 mm), dengan efisiensi waktu lebih cepat dan dengan generasi yang jauh lebih sedikit dari Simple GA. Pada level 2 implementasi pada lingkungan Gazebo berhasil menyelesaikan 5 titik uji dan menjalankan lintasan U dan O, mencatat mean error trajectory sebesar 0,0748 mm dan 0,0783 mm. Pada Level 3 di lingkungan real plant , robot OpenManipulator-X berhasil menyelesaikan 5 titik uji dan menghasilkan mean error 0,7772 mm untuk lintasan -U serta mean error 0,7240 mm untuk lintasan-O. Faktor degradasi dari pengujian level 1 ke level 2 baik uji titik tunggal maupun lintasan hampir tidak menambah error. Sedangkan, faktor degradasi dari pengujian level 1 ke level 3 untuk uji titik tunggal sebesar 6.48 hingga 6.51 kali dan untuk pengujian lintasan adalah sebesar 12,51 h ingga 12,91 kali. Kerangka kerja hibrida NN-GA ini berpotensi untuk diterapkan pada platform robot manipulator lain.
===============================================================================================================================
This research proposes a Hybrid Neural Network–Genetic Algorithm (NN-GA) system to accurately solve the Inverse Kinematics problem, using the 4-DoF OpenManipulator-X robot which is redundant with respect to position tasks as the case study. The proposed hybrid system integrates the computational speed of Neural Networks (NN) with the accuracy of Genetic Algorithms (GA). In Stage I, a GA Tuner process automatically optimizes the hyperparameters of three specialist NN models. In Stage II, the parallel predictions of the three NNs are selected through an Anchor Selection mechanism (selecting the best NN solution) based on Forward kinematics (FK) evaluation, then refined by GA Refinement until sub-millimeter target accuracy is achieved. The system was validated across three tiered levels. At Level 1, pure testing of the hybrid system's predictions achieved a 100% Success Rate (SR) on 5 Single Points (mean error of 0.0891 mm) and 95.80% SR on 500 random points. A comparative study on 20 points showed that the Hybrid NN-GA produced a mean error of 0.0769 mm and 100% SR, outperforming NN-Only (mean error of 59.54 mm) and Simple GA (mean error of 0.67 mm), with faster computation time and significantly fewer generations than Simple GA. At Level 2, implementation in the Gazebo environment successfully completed 5 test points and executed U- and O-shaped trajectories, recording mean trajectory errors of 0.0748 mm and 0.0783 mm. At Level 3, in the real-plant environment, the OpenManipulator-X robot successfully completed 5 test points, yielding mean errors of 0.7772 mm for the U-trajectory and 0.7240 mm for the O-trajectory. The degradation factor from Level 1 to Level 2 testing, for both single-point and trajectory tests, added almost no additional error. Meanwhile, the degradation factor from Level 1 to Level 3 testing was 6.48–6.51 times for single-point tests and 12.51–12.91 times for trajectory tests. This NN-GA hybrid framework has the potential to be applied to other manipulator robot platforms
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | Genetic Algorithm, Kinematika Balik, Neural Network, OpenManipulator-X, Robot Redundan,Sistem Hibrida Genetic Algorithm, Hybrid System, Inverse Kinematics, Neural Network, OpenManipulator-X, Redundant Robot |
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK6592.A9 Automatic tracking. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis |
| Depositing User: | Muhammad Lukmanul Hakim Hidayatullah |
| Date Deposited: | 30 Jul 2026 00:56 |
| Last Modified: | 30 Jul 2026 00:56 |
| URI: | http://repository.its.ac.id/id/eprint/139087 |
Actions (login required)
 |
View Item |