Interaksi Robot Manusia Berbasis Large Language Model Untuk Pembuatan Movement Plan Pada Robot Quadruped-Legged

Mahendra, Vincentius Gusti Putu Agung Bagus (2025) Interaksi Robot Manusia Berbasis Large Language Model Untuk Pembuatan Movement Plan Pada Robot Quadruped-Legged. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024211071-Undergradute_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (19MB) | Request a copy

Abstract

Penelitian ini mengusulkan sistem interaksi manusia–robot berbasis model bahasa besar (LLM) untuk menghasilkan rencana gerak (movement plan) pada robot berkaki empat (quadruped-legged) Jueying Lite 3. Sistem ini terdiri dari tiga komponen utama: pemetaan dan lokalisasi indoor menggunakan LiDAR Leishen C16 dengan metode SLAM berbasis Fast-LIO, pemrosesan perintah bahasa alami berbahasa Indonesia melalui Google Gemini-1.5-Flash pada Google Vertex AI, serta perencanaan dan eksekusi navigasi end-to-end menggunakan Robot Operating System (ROS) dan navigation stack. Proses dimulai dengan pembangunan peta digital lingkungan Tower 2 Lantai 9 ITS, dilanjutkan dengan penerjemahan perintah teks menjadi rangkaian aksi oleh LLM, yang kemudian dikonversi menjadi waypoint untuk dieksekusi oleh robot secara otomatis. Hasil menunjukkan bahwa model bahasa besar berhasil menyusun rencana aksi dengan akurasi 84%, sementara sistem navigasi robot mampu mengeksekusi pergerakan sesuai rencana dengan tingkat keberhasilan 96,5%. Beberapa kegagalan navigasi lokal terjadi akibat fluktuasi posisi point cloud dan ketidakstabilan robot saat melewati area sempit. Secara keseluruhan, sistem ini menunjukkan potensi besar dalam meningkatkan interaksi natural antara manusia dan robot, serta memperkuat kemampuan otonom robot berkaki empat dalam menjalankan tugas navigasi. Pengembangan lebih lanjut direkomendasikan untuk meningkatkan stabilitas gerak, keandalan sistem dalam kondisi ekstrem, dan integrasi modalitas suara guna mendukung interaksi dua arah secara lebih efektif.
======================================================================================================================================
This research proposes a human–robot interaction system based on a large language model (LLM) to generate movement plans for the quadruped-legged robot Jueying Lite 3. The system comprises three main components: indoor mapping and localization utilizing the Leishen C16 LiDAR with Fast-LIO-based SLAM, natural language command processing in Indonesian via Google Gemini-1.5-Flash on Google Vertex AI, and end-to-end navigation planning and execution using the Robot Operating System (ROS) and its navigation stack. The process begins with the construction of a digital map of the 9th floor of Tower 2 at ITS, followed by the conversion of text prompts into a sequence of actions by the LLM, which are then translated into waypoints for autonomous execution by the robot. Results show that the language model successfully generates action plans with an accuracy of 84%, while the robot navigation system executes movement plans with a success rate of 96.5%. Some local navigation failures occurred due to fluctuations and instability in the point cloud when the robot traversed narrow areas. Overall, the system demonstrates strong potential for enhancing natural human–robot interaction and improving autonomous navigation capabilities in quadruped-legged robots. Future development is recommended to improve movement stability, increase system reliability in challenging environments, and integrate voice-based modalities to support more interactive and responsive

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Human–Robot Interaction, Large Language Model, Quadruped-Legged Robot, SLAM, Navigasi Otonom, Human–Robot Interaction, Large Language Model, Quadruped-Legged Robot, SLAM, Autonomous Navigation
Subjects:	T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TJ Mechanical engineering and machinery > TJ211 Robotics. T Technology > TJ Mechanical engineering and machinery > TJ211.415 Mobile robots T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7895.S65 Speech recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Vincentius Gusti Putu Agung Bagus Mahendra
Date Deposited:	30 Jul 2025 06:12
Last Modified:	30 Jul 2025 06:12
URI:	http://repository.its.ac.id/id/eprint/123197

Actions (login required)

View Item