Home > Published Issues > 2020 > Volume 9, No. 9, September 2020 >

Autonomous Searching Robot with Object Recognition Based on Neural Networks

Jakub Czygier, Przemysław Dąbrowski, Robert Grabowy, Maciej Rećko, and Kazimierz Dzierżek
Bialystok University of Technology, Department of Automatic Control and Robotics, Faculty of Mechanical Engineering, Białystok, Poland

Abstract—The presented paper describes the design and implementation process of a six-wheeled autonomous robotic platform. The device is equipped with advanced object recognition algorithms based on neural networks. It is low cost and easy to build a device capable of roving in varied terrain. It is ideal for scouting missions, or mapping visited areas. Following a brief introduction into the explored topic, a design process of a considered robot. Mechanical, electronic and onboard sensory systems are found in the next part of the article. The following section consists of a description of the control and navigation approach used in our system. Then, YOLO object recognition algorithm is explained, followed by a proposed experiment to validate its abilities. Set of tasks were given to the robot, that had to complete them autonomously. Results are highly satisfying. The YOLO algorithm proved useful in object recognition providing crucial data required during autonomous drive mode.

Index Terms—mobile robots, autonomous robots, object recognition

Cite: Jakub Czygier, Przemysław Dąbrowski, Robert Grabowy, Maciej Rećko, and Kazimierz Dzierżek, "Autonomous Searching Robot with Object Recognition Based on Neural Networks," International Journal of Mechanical Engineering and Robotics Research, Vol. 9, No. 9, pp. 1347-1352, September 2020. DOI: 10.18178/ijmerr.9.9.1347-1352