Çok Kameralı Görü Tabanlı Mobil Robot Kontrolü ve Yol Planlaması

Abstract

Visual-based controlling concerns how to manage a robot by using image-based features obtained from imaging devices. There are two types of configurations; internal vision configuration -IVC (the camera is in the device) and external vision configuration -EVC (the camera is out of the device). A multi-camera process with EVC has been employed to generate input for modeling behaviors of a mobile robot in this study. Gaussian-based controller and adaptive artificial potential field (A-APF) have been utilized for control and path planning processes, respectively. Four webcams with the same specifications are operated from the same height to control the mobile robot within a specific area. Images are stitched by using identic features of intersection areas in the first stage. Then, the color-based object detection is performed on this image to detect positions and other information related to the target, robot ,and obstacles. In the next step, to obtain a feasible and safe path plan between target and robot, an adaptive potential field algorithm is performed. Then, the controller is used to model the robot motions according to the path plan. In each control iteration, only one camera is activated according to the local position obtained from the global position of the moving robot. The simulation and physical experiment results demonstrate that the multi-camera configuration provides good performance and efficiency.