Implementasi Model Predictive Control dan Whole-Body Impulse Control pada Quadruped Robot Dog

Maulana, Figo Arzaki (2026) Implementasi Model Predictive Control dan Whole-Body Impulse Control pada Quadruped Robot Dog. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5022221041-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Robot anjing quadruped memiliki potensi besar untuk menavigasi medan tidak terstruktur pada berbagai aplikasi, seperti pencarian dan penyelamatan, inspeksi industri, serta eksplorasi lingkungan berbahaya. Namun, perancangan sistem kontrol yang mampu menghasilkan lokomosi dinamis, dan robust terhadap berbagai medan masih menjadi tantangan. Di Indonesia, penelitian mengenai lokomosi berbasis model pada robot anjing quadruped masih relatif terbatas, baik dari sisi pendekatan teori maupun implementasi pada platform lokal. Penelitian ini bertujuan mengimplementasikan dan mengadaptasi arsitektur kontrol berbasis model yang menggabungkan Model Predictive Control (MPC) dan Whole-Body Impulse Control (WBIC) pada robot quadruped lokal bernama Fiammetta. Sistem yang dikembangkan seluruh arsitektur kontrol seperti estimasi state, pembangkit gait, terutama MPC sebagai penghasil gaya reaksi tanah optimal dalam jangka waktu lebih panjang dengan model simpel, serta WBIC sebagai pengendali yang merekonsiliasi hasil dari MPC dengan kinematika dan model dinamika komplit untuk menghasilkan perintah posisi, kecepatan dan torsi sendi. Kinerja sistem dievaluasi dalam lingkungan simulasi MuJoCo melalui empat skenario pengujian, yaitu medan datar, ramp, tangga, dan gangguan eksternal. Selain itu, dilakukan perbandingan terhadap dua metode kontrol alternatif, yaitu kontrol berbasis Virtual Model Control (PD-VMC) dan kontrol Inverse Kinematics murni (IK). Penelitian ini diharapkan dapat menjadi referensi teknis dan fondasi bagi pengembangan robot quadruped dengan lokomosi dinamis di Indonesia. Hasil pengujian menunjukkan bahwa MPC-WBIC menghasilkan pelacakan kecepatan longitudinal dan Cost of Transport terbaik pada seluruh variasi kecepatan yang diuji serta juga dapat mencapai kecepatan aktual tertinggi pada medan datar. Pada pengujian ramp, MPC-WBIC mempertahankan tingkat keberhasilan traversal sebesar 100% hingga kemiringan 30°, sedangkan kontroler IK gagal pada kemiringan 25° dan 30°. Pada pengujian tangga, MPC-WBIC berhasil menyelesaikan seluruh konfigurasi tinggi anak tangga hingga 18 cm dengan tingkat keberhasilan 100%, sementara PD-VMC hanya berhasil pada tangga setinggi 10 cm dan IK gagal pada seluruh konfigurasi. Pada pengujian gangguan eksternal, MPC-WBIC menunjukkan tingkat ketahanan tertinggi dengan keberhasilan 100% pada hampir seluruh konfigurasi gangguan yang diuji. Secara keseluruhan, hasil penelitian menunjukkan bahwa kombinasi MPC dan WBIC memberikan performa lokomosi yang lebih baik dibandingkan PD-VMC dan IK dalam aspek pelacakan kecepatan, stabilitas, efisiensi energi, serta ketahanan terhadap variasi medan dan gangguan eksternal.
=================================================================================================================================
Quadruped robot dogs have great potential for navigating unstructured terrain in various applications, such as search and rescue, industrial inspection, and exploration of hazardous environments. However, designing a control system capable of producing dynamic and robust locomotion across diverse terrains remains a challenge. In Indonesia, research on model-based locomotion of quadruped robot dogs is still relatively limited, both in terms of theoretical approaches and implementation on local platforms. This research aims to implement and adapt a model-based control architecture that combines Model Predictive Control (MPC) and Whole-Body Impulse Control (WBIC) on a local quadruped robot named Fiammetta. The developed system encompasses the entire control architecture, including state estimation, gait generation, and most importantly MPC, which generates optimal ground reaction forces over a longer time horizon using a simplified model, as well as WBIC, a controller that reconciles the MPC outputs with the kinematics and full dynamics model to produce joint position, velocity, and torque commands. The system performance is evaluated in the MuJoCo simulation environment through four test scenarios: flat terrain, ramps, stairs, and external disturbances. In addition, comparisons are made against two alternative control methods, namely a control based on Virtual Model Control (PD-VMC) and a pure Inverse Kinematics (IK) control. This research is expected to serve as a technical reference and foundation for the development of quadruped robots with dynamic locomotion in Indonesia. The test results show that MPC-WBIC yields the best longitudinal velocity tracking and Cost of Transport across all tested speed variations, and also achieves the highest actual speed on flat terrain. In the ramp tests, MPC-WBIC maintains a 100% traversal success rate up to a 30° incline, while the IK controller fails at 25° and 30° inclines. In the stair tests, MPC-WBIC successfully completes all step height configurations up to 18 cm with a 100% success rate, whereas PD-VMC only succeeds on stairs as high as 10 cm and IK fails on all configurations. In the external disturbance tests, MPC-WBIC exhibits the highest robustness, achieving 100% success on nearly all tested disturbance configurations. Overall, the results of this research demonstrate that the combination of MPC and WBIC delivers better locomotion performance compared to PD-VMC and IK in terms of velocity tracking, stability, energy efficiency, and robustness to terrain variations and external disturbances.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Robot Anjing Quadruped, Lokomosi Quadruped, Model Predictive Control, Whole-Body Impulse Control
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ211 Robotics. T Technology > TJ Mechanical engineering and machinery > TJ211.4 Robot motion
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Figo Arzaki Maulana
Date Deposited:	21 Jul 2026 04:39
Last Modified:	21 Jul 2026 07:38
URI:	http://repository.its.ac.id/id/eprint/135885

Actions (login required)

View Item