Article Per Year (5 Year)
p-Index From 2020 - 2025
1.8
P-Index
This Author published in this journals
All Journal ENGINEERING Jurnal Bioterdidik: Wahana Ekspresi Ilmiah Unes Law Review Budapest International Research and Critics Institute-Journal (BIRCI-Journal): Humanities and Social Sciences Journal Presumption of Law Jurnal Ilmiah Wahana Pendidikan Jurnal Riset Rumpun Agama dan Filsafat AN-NIDA: Jurnal Pendidikan Islam Inovate : Jurnal Ilmiah Inovasi Teknologi Informasi Jurnal Idaaratul Ulum Indonesian Journal of Aerospace
Imron Rosadi
Unknown Affiliation
Religion Aerospace Engineering Agriculture, Biological Sciences & Forestry Arts Humanities Astronomy Computer Science & IT Economics, Econometrics & Finance Education Engineering Law, Crime, Criminology & Criminal Justice Physics Public Health Social Sciences Other

Published : 12 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Indonesian Journal of Aerospace

 1 
Optimization of Powered Landing Control for Reusable Rockets Using Softmax DDGN Rafika Arum Sari; Muhammad Hadi Widanto; Imron Rosadi
Indonesian Journal of Aerospace Vol. 22 No. 2 (2024): Indonesian Journal Of Aerospace
Publisher : BRIN Publishing

Show Abstract | Download Original | Original Source | Check in Google Scholar

Abstract

This research presented a novel approach to optimize powered landing control on reusable rockets by using Softmax Double Deep Q-Networks (DDQN). We combined the advantages of Double DQN with Softmax exploration and curriculum learning to achieve precise and efficient landing control. Through extensive experiments in a specially developed 2D simulation environment, our method achieves improved landing accuracy by 37% (reduced final position error from 2.4 m to 1.5 m), better fuel efficiency by 28% (reduced average fuel consumption from 850 kg to 612 kg per landing), and improved adaptability to initial conditions (improved successful landing rate from 76% to 94% across a wide range of altitudes and initial orientations) compared to traditional PID control methods. The results showed that the curriculum learning method significantly outperformed the non-curriculum approach, achieving 27% higher average awards (11.97 vs. 8.61) and 60% better performance consistency as measured by standard deviation (0.92 vs. 2.29). Both Softmax and ε-greedy exploration strategies proved effective with curriculum learning, with ε-greedy DDQN achieving the highest average award of 11.97. This approach allows for higher precision rocket landings while reducing operational costs through.
Co-Authors ADHITYA WIDYA KARTIKA Agus Siswanto Agus Wibowo Ahmad Farid Aisyah Nurhayati Akhsanul Huda amat chaeroni Andi Abdul Hanafi Andi Rosa Andri Nirwana AN Arifin, Hafizh BAMBANG SUJATMIKO Berti Yolida Desi Fitriani Ginanjar Setyo Permadi Irsyad Khoerul Imam M. Zaky Hamdie Moh Abdul Kholiq Hasan Muhammad Hadi Widanto Nurikhsan, Muhamad Alfan Rafika Arum Sari Reza Deniarti Rini Indah Sari Rohmah, Lusy Meilani Nur’aeni Saiful Akhyar Lubis Siti Waspiatul Kamilah Tanhella Zein Vitadiar Tazkia Nur Fateha Tri Jalmo