Nugroho, Arif (2024) Modeling of Robotic Motion Control for Consensus of a Multi-Agent System Based on Neural Network Architecture: Case Study on NAO Robot Arms. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 07111860010014-Dissertation.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
07111860010014-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 July 2026. Download (6MB) | Request a copy |
Abstract
Robot artikulasi, seperti lengan robot, telah menarik perhatian luas dalam aplikasi robot modern karena fleksibilitas dan keserbagunaannya. Struktur kinematika mereka, yang berisi sekumpulan benda kaku yang dihubungkan oleh sambungan berputar, memungkinkan mereka memiliki lebih banyak fleksibilitas dalam gerakan. Kemampuannya dalam melakukan berbagai macam gerakan juga menjadikannya serbaguna. Kehadiran beberapa sambungan putar pada lengan robot NAO memungkinkan mereka melakukan berbagai macam gerakan namun, pada saat yang sama, membuatnya lebih rumit untuk dikendalikan. Peningkatan dimensi ruang kontrol karena adanya beberapa sambungan putar membuat lengan robot NAO memerlukan lebih banyak tindakan kontrol untuk melakukan gerakan yang diinginkan. Untuk mengeksekusi gerakan yang diinginkan, kita perlu merancang kontrol gerak robot. Dalam penelitian disertasi ini, rancangan kendali gerak robot akan secara khusus ditujukan pada konsensus sistem multi-agen. Konsensus dalam sistem multi-agen adalah bagian dari sistem multi-agen kooperatif yang secara khusus berfokus pada pengendalian sekumpulan agen untuk mencapai keadaan bersama. Sistem multi-agen biasanya berisi sekumpulan agen yang keadaan awalnya berbeda satu sama lain. Dalam konsensus sistem multi-agen, setiap agen diharapkan menyesuaikan keadaannya sendiri sesuai dengan keadaan informasi yang diterima dari agen tetangga sehingga keadaan gerak semua agen menyatu ke keadaan yang sama. Untuk melakukan pelaksanaan tugas secara paralel dalam sistem yang terdiri dari lengan robot, perlu dirancang kontrol gerak berbasis konsensus. Kontrol gerak berbasis konsensus yang kami rancang dalam penelitian ini digunakan untuk menyinkronkan keadaan gerak lengan robot NAO. Tujuan dari penelitian ini adalah untuk memodelkan kontrol gerak berbasis konsensus pada kasus lengan robot NAO. Untuk mencapai tujuan penelitian ini, ada langkah-langkah berurutan yang harus dilakukan dalam penelitian ini. Langkah-langkah berurutan ini dapat dirinci sebagai berikut: (1) mengidentifikasi struktur kinematika lengan robot NAO, (2) memodelkan kinematika maju lengan robot NAO, (3) membuat kumpulan data untuk membangun kinematika invers berbasis jaringan saraf , (4) mengembangkan model kinematika invers berbasis jaringan saraf untuk lengan robot NAO, (5) merumuskan aturan pembaruan konsensus berdasarkan skema konsensus sistem multi-agen, dan (6) merancang kontrol gerak berbasis konsensus baik di ruang bersama maupun ruang operasional. Dalam perancangan kendali gerak berbasis konsensus, kami berhasil merumuskan model kinematika maju untuk lengan robot NAO dan model kinematika maju ini telah diverifikasi dengan benar. Dengan menggunakan model kinematika maju yang terverifikasi ini, kami menghasilkan kumpulan data untuk membangun kinematika invers berbasis jaringan saraf dan kumpulan data ini telah disebarluaskan di repositori data publik: IEEE Dataport dan data Mendeley. Dengan menggunakan dataset yang kami buat dalam penelitian ini, kami mengembangkan model kinematika invers berbasis jaringan saraf untuk lengan robot NAO. Dalam penelitian ini, kami juga mengusulkan pendekatan baru untuk meningkatkan kinerjanya dengan cara mengintegrasikan kinematika invers berbasis jaringan saraf dengan Jacobian. Dalam pendekatan yang diusulkan ini, kinematika invers berbasis jaringan saraf berfungsi sebagai kontrol umpan maju, dan Jacobian bertindak sebagai kontrol umpan balik. Kehadiran Jacobian membantu meminimalkan kesalahan yang tersisa dari kinematika invers berbasis jaringan saraf, sehingga memberikan kinerja yang lebih baik. Sebagai buktinya, hasil perbandingan menunjukkan bahwa rata-rata MSE untuk optimasi gerombolan partikel (PSO) adalah 3,47 x 10-3 rad, 1,19 x 10-3 rad untuk jaringan saraf, dan kemudian 3,72 x 10-5 rad untuk pendekatan yang kami usulkan. . Hasil perbandingan kinerja menunjukkan bahwa pendekatan yang kami usulkan menghasilkan rata-rata UMK terendah dibandingkan pendekatan lainnya. Selain itu, dalam penelitian ini kami juga berhasil merumuskan aturan pemutakhiran konsensus baik dalam skema konsensus tanpa pemimpin maupun skema konsensus mengikuti pemimpin. Aturan pembaruan konsensus dalam kasus keadaan yang berubah terhadap waktu lebih dapat diterapkan dalam berbagai aplikasi dibandingkan dengan keadaan yang tidak berubah terhadap waktu karena sifat dari hampir semua sistem yang berubah secara dinamis seiring waktu. Karena temuan yang disebutkan di atas, kami telah berhasil mengembangkan kontrol gerak berbasis konsensus untuk lengan robot NAO. Dengan menerapkan kontrol gerak berbasis konsensus yang dirancang, perbedaan keadaan gerak di antara lengan robot dapat diatasi sehingga pelaksanaan tugas secara paralel dapat dicapai
============================================================================================================================
Articulated robots, such as robotic arms, have attracted widespread attention in modern robot applications due to their flexibility and versatility. Their kinematics structure, which contains a set of rigid bodies connected by revolute joints, allows them to have more flexibility in the movement. Their capability to perform various kinds of motions also makes them versatile. The presence of multiple revolute joints in the NAO robot arms enables them to perform a wide range of motions but, at the same time, makes them more complex to control. The increased dimensionality of the control space due to existing multiple revolute joints makes the NAO robot arms require more control actions in the case of executing desired motions. To execute the desired motions, we need to design the robotic motion control. In the research of this dissertation, the designed robot motion control will be particularly addressed on the consensus of a multi-agent systems. The consensus in a multi-agent system is a subset of cooperative multi-agent systems that specifically focus on controlling a set of agents to reach a common state. The multi-agent system typically contains a set of agents whose initial states are different from one another. In the consensus of a multi-agent system, each agent is supposed to adjust its own states in accordance with the information states received from the neighbor agents so that the motion states of all agents converge to a common state. To perform the parallel execution of tasks in a system that consists of robotic arms, it is necessary to design the consensus-based motion control. The consensus-based motion control that we design in this research is employed to synchronize the motion states of the NAO robot arms. The objective of this research is to model the consensus-based motion control in the case of NAO robot arms. To achieve the objective of this research, there are sequential steps that must be conducted in this research. These sequential steps can be listed as follows: (1) identifying the kinematics structure of the NAO robot arms, (2) modeling the forward kinematics of the NAO robot arms, (3) creating the dataset to build the neural network-based inverse kinematics, (4) developing the neural network-based inverse kinematics model for the NAO robot arms, (5) formulating the consensus update rule based on the consensus scheme of a multi-agent system, and (6) designing the consensus-based motion control in both the joint space and the operational space. In the design of consensus-based motion control, we successfully formulated the forward kinematics model for the NAO robot arms and this forward kinematics model has been verified correctly. By using this verified forward kinematics model, we generated the dataset to build the neural network-based inverse kinematics and this dataset has been disseminated in public data repositories: IEEE Dataport and Mendeley data. By using the dataset that we created in this research, we developed the neural network-based inverse kinematics model for the NAO robot arms. In this research, we also proposed a novel approach to improve its performance by means of integrating the neural network-based inverse kinematics with the Jacobian. In this proposed approach, the neural network-based inverse kinematics serves as the feedforward control, and the Jacobian acts as the feedback control. The presence of the Jacobian helpfully minimized the remaining error of the neural network-based inverse kinematics, thereby providing better performance. As proof, the comparison result showed that the averaged MSE for the particle swarm optimization (PSO) was 3.47 x 10-3 rad, 1.19 x 10-3 rad for the neural network, and then 3.72 x 10-5 rad for our proposed approach. The performance comparison result indicated that our proposed approach produces the lowest averaged MSE than the other ones. Besides, in this research, we also successfully formulated the consensus update rule in both the leaderless consensus scheme and the leader-following consensus scheme. The consensus update rule in the case of time-varying states is more applicable in a wide range of applications than that in the time-invariant states due to the nature of almost all systems that dynamically change over time. Because of the aforementioned findings, we have successfully developed the consensus-based motion control for the NAO robot arms. By applying the designed consensus-based motion control, the difference in the motion states among the robotic arms can be overcome such that the parallel execution of tasks can be achieved
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Kontrol gerak robot, Konsensus sistem multi-agen, lengan robot NAO, Eksekusi tugas paralel, Jaringan saraf; Robotic motion control, Consensus of a multi-agent system, NAO robot arms, Parallel execution of tasks, Neural network |
| 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 > 20001-(S3) PhD Thesis |
| Depositing User: | Arif Nugroho |
| Date Deposited: | 19 Feb 2024 01:42 |
| Last Modified: | 19 Feb 2024 01:42 |
| URI: | http://repository.its.ac.id/id/eprint/107357 |
Actions (login required)
 |
View Item |