Autonomous Utility Vehicle Development by Using Depth Camera and Machine Learning Conference

Perez-Perez, D, Gonzalez, G, Tansel, I. (2023). Autonomous Utility Vehicle Development by Using Depth Camera and Machine Learning . 2023-March 128-132. 10.54808/IMCIC2023.01.128

cited authors

  • Perez-Perez, D; Gonzalez, G; Tansel, I

authors

abstract

  • Machine vision is one of the most important components in autonomous systems. Autonomous vehicles process sensor data to identify objects, estimate their locations, and determine the proper path. Development of a multipurpose program is complex, expensive, takes ample time, and may have many bugs. In this study, Intel Realsense D455 depth camera was used to develop a navigation system for small utility cars. The camera was integrated with Robot Operating System (ROS) for data analysis and decision making. The D455 obtained images and calculated the distance of every pixel in the frame. A deep-learning neural network with custom-trained data was utilized to detect a test target's bounding box. The obstacle avoidance algorithm detected objects' distances on the path from the depth information provided by the D455. The machine vision system calculated the proper path and operated the DC motors to drive the vehicle towards the target. The developed multipurpose machine vision system for utility vehicles was adopted to the Autonomous Asphalt Laying Machine (AALaM). In addition to the navigation, the asphalt feed and delivery components were also integrated to the ROS and the entire system worked successfully.

publication date

  • January 1, 2023

Digital Object Identifier (DOI)

International Standard Book Number (ISBN) 13

start page

  • 128

end page

  • 132

volume

  • 2023-March