Navigation of autonomous mobile robots in dynamic unknown environments based on dueling double deep q networks
Year: 2024
Type: article
Abstract: In recent years, the use of artificial intelligence applications has increased due to developments in hardware. These applications are used in many fields, including robotics. Deep reinforcement learning, one of the subfields of artificial intelligence, is among the topics that researchers focus on. Deep reinforcement learning has a structure similar to that of human learning. This makes it suitable for use in the navigation problems of robots in unknown environments. This study focuses on the use of the Dueling Double Q Networks algorithm, one of the deep reinforcement learning algorithms, in robot navigation problems. The goal of the study is to develop a network model that can assist a robot in navigating an unknown environment, avoiding obstacles, detecting targets, and reaching those targets. Three network models based on the Dueling Double Q Networks algorithm were designed and trained using depth images from a depth camera and direction and distance information from a camera. The trained models were tested in both simple and complex test environments. Success in reaching the target and the total reward amount obtained by the model were used as success metrics. The results showed that the most successful model was the Dueling Double Q Networks-C model. This model achieved a success rate of 89% in a simple test environment and 87% in a complex test environment. Experimental results demonstrate the potential for deep reinforcement learning algorithms, particularly the Dueling Double Q Networks algorithm, to be effectively used in the navigation problems of robotics.
Authors: Koray Ozdemir, Adem Tuncer
Institution: Yalova University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $3,270
Funder: Yalova Üniversitesi
Award: G5256409681
OpenAlex ID: W4403824395