DEEP LEARNING-BASED VISUAL SERVOING ALGORITHM FOR WHEELED MOBILE ROBOT CONTROL
Aleksandar Jokić, Đorđe Jevtić, Katarina Brenjo, Milica Petrović, Zoran Miljković · 2024
Production-oriented companies that aspire to the concept of Industry 4.0 primarily focus on the increasing flexibility and reconfigurability of the entire manufacturing system.By integrating a robotic material transport/handling system that features a high level of efficiency, flexibility, and intelligence, the entire manufacturing system reaps the benefits.With that in mind, the authors propose a deep learning-based visual servoing algorithm for the intelligent control of a wheeled mobile robot.By utilizing a visual servoing algorithm, mobile robotic systems can flexibly and efficiently adapt their trajectories to real-world conditions.Moreover, deep learning algorithms allow mobile robots to learn robust visual features that make visual servoing even more applicable.The authors utilize state-of-the-art deep learning models to train the mobile robot to perform visual servoing even without distinct features that are necessary for such a system to function properly.Experimental evaluation with the own developed mobile robot RAICO -Robot with Artificial Intelligence based COgnition has shown the benefits of the proposed visual control algorithm.