Pengembangan Navigasi Robot Nonholonomic Menggunakan Self-Adaptive Dynamic Window Approach Berbasis Fuzzy Logic

Purwanto, Agung (2026) Pengembangan Navigasi Robot Nonholonomic Menggunakan Self-Adaptive Dynamic Window Approach Berbasis Fuzzy Logic. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6022232002-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (24MB) | Request a copy

Abstract

Robot mobile otonom semakin banyak digunakan di lingkungan indoor seperti restoran, rumah sakit, dan gudang untuk mendukung tugas pengantaran yang menuntut robot menavigasi ruang berrintangan secara aman sekaligus memenuhi tenggat waktu misi. Dynamic Window Approach (DWA) merupakan salah satu algoritma navigasi lokal yang paling banyak digunakan pada robot nonholonomic, namun memiliki dua keterbatasan mendasar: bobot fungsi evaluasinya bersifat tetap sehingga tidak dapat menyesuaikan diri terhadap perubahan kondisi lingkungan, dan algoritma ini tidak memiliki kesadaran temporal sehingga tidak dapat memodulasi kecepatan pada kondisi waktu yang semakin menipis. Penelitian ini mengusulkan Self-Adaptive Dynamic Window Approach (SA-DWA) berbasis fuzzy logic yang mengatasi kedua keterbatasan tersebut melalui mekanisme adaptasi tiga lapis. Bobot arah diadaptasi terhadap jarak rintangan terdekat melalui fungsi logaritmik; bobot kesetiaan jalur dan bobot kecepatan dihasilkan sistem inferensi fuzzy Mamdani 25 aturan yang dikalibrasi dari 483 sampel empiris di lingkungan nyata; sedangkan tekanan waktu dikelola melalui Time Pressure Ratio (TPR) yang menggeser batas kecepatan atas dynamic window secara piecewise-quadratic. Kontribusi utama penelitian ini adalah penempatan informasi temporal pada level batas kecepatan, bukan di dalam fungsi penilaian lintasan, sehingga urgensi waktu hanya memperluas ruang pilihan kecepatan tanpa menggangu komponen keselamatan dalam pemilihan trajektori. SA-DWA dipadukan dengan Kanayama Kinematic Controller sebagai tracking kecepatan referensi. Pengujian dilakukan melalui simulasi ROS2 Jazzy/Gazebo dan eksperimen pada robot fisik TurtleBot4 Standard, mencakup lima skenario di lingkungan statis dan dinamis pada dua tingkat tekanan waktu. SA-DWA secara konsisten menempuh misi lebih cepat daripada DWA konvensional dan Improved DWA, dengan keunggulan 0,9%–3,7% pada lingkungan statis dan hingga 10,8% pada lingkungan dinamis, serta 7,86% lebih cepat pada pengujian fisik, dengan tingkat keberhasilan 100% dan tanpa tabrakan pada seluruh kombinasi pengujian. Analisis lebih lanjut menunjukkan bahwa mekanisme TPR memperbaiki waktu tempuh secara langsung. Biaya peningkatan kecepatan ini muncul sebagai kenaikan jerk di lingkungan terbuka, namun di lingkungan dinamis kecepatan yang lebih tinggi justru menstabilkan profil gerak dan menurunkan jerk. Jarak aman (clearance) terjaga konstan saat kondisi beralih dari tekanan waktu rendah ke kritis, membuktikan secara empiris bahwa efisiensi temporal dan keselamatan spasial beroperasi secara independen.
=================================================================================================================================
Autonomous mobile robots are increasingly deployed in indoor environments such as restaurants, hospitals, and warehouses to support delivery tasks that require the robot to navigate obstacle-filled spaces safely while meeting mission deadlines. The Dynamic Window Approach (DWA) is one of the most widely used local navigation algorithms for nonholonomic robots, yet it has two fundamental limitations: its evaluation weights are fixed and cannot adapt to changing environmental conditions, and it has no temporal awareness, thus cannot modulate its velocity under conditions of dwindling time. This research proposes a Self-Adaptive Dynamic Window Approach (SA-DWA) based on fuzzy logic that addresses both limitations through a three-layer adaptation mechanism. The heading weight is adapted to the nearest-obstacle distance using a logarithmic function; the path-following weight and velocity weight are produced by a 25-rule Mamdani fuzzy inference system calibrated from 483 empirical samples collected in a real environment; and time pressure is managed through the Time Pressure Ratio (TPR), which shifts the upper velocity limit of the dynamic window in a piecewise-quadratic manner. The main contribution is placing the temporal information at the velocity-limit level rather than inside the trajectory scoring function, so that temporal urgency only widens the set of available velocities without ever competing with the safety component during trajectory selection. SA-DWA is paired with the Kanayama Kinematic Tracking Controller to track the reference velocity. Testing was conducted through ROS2 Jazzy/Gazebo simulation and physical experiments on a TurtleBot4 Standard robot, covering five scenarios across static and dynamic environments under two levels of time pressure. SA-DWA consistently completed missions faster than conventional DWA and Improved DWA, with advantages of 0.9%-3.7% in static environments and up to 10.8% in dynamic environments, and 7.86% faster in physical testing, with a 100% success rate and zero collisions across all combinations. Further analysis shows that the TPR mechanism improves travel time directly.. The cost of this higher speed appears as increased jerk in open environments, yet in dynamic environments the higher speed instead stabilizes the motion profile and reduces jerk. The safety clearance remains constant when conditions shift from low to critical time pressure, empirically demonstrating that temporal efficiency and spatial safety operate independently

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Dynamic Window Approach, Self-Adaptive, Mamdani Fuzzy Logic, Time Pressure Ratio, Nonholonomic Robot, ROS2, TurtleBot4
Subjects:	T Technology > TJ Mechanical engineering and machinery > TJ211 Robotics. T Technology > TJ Mechanical engineering and machinery > TJ211.415 Mobile robots
Divisions:	Faculty of Electrical Technology > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	Agung Purwanto
Date Deposited:	30 Jul 2026 02:17
Last Modified:	30 Jul 2026 02:17
URI:	http://repository.its.ac.id/id/eprint/139424

Actions (login required)

View Item