Vision based and real time obstacle avoidance of mobile robot

In the navigation of mobile robots, target recognition and obstacle avoidance are one of the most important research topics studied. Very complex algorithms and operations are required for autonomous robots to recognize targets and move towards a specific target. In this regard, various mapping methods, sensor data and various approaches have been adopted and designs have been made. The aim of this study is to realize an autonomous robot simulation that can move without hitting obstacles with computerized vision without using a map and using any sensor data. In the study, a two-step process is followed for the traveling robot to navigate avoiding obstacles. In the first stage, the images taken from the robot camera are processed and ground separation is made in the image. In the second stage, the area marked as the floor is checked. If it is suitable for the robot to move forward, forward motion is made. If it is not suitable for movement, that is, if any obstacle is detected, the existing image is reprocessed, making only a rotational movement to the right or left. Since all these processes continue recursively if there is a similar situation in the next image, it continues to rotate or forward motion is started. The system, consisting of two stages, operates in real-time. The biggest advantage of this is that the system is useful in dynamic environments. The environment is evaluated for each image taken on the webcam, and the robot can decide on movements such as rotation or progression. In this study, the ROS operating system designed for robots and consisting of open sources is used. Gazebo, which can be integrated with ROS, was chosen as the simulation medium. Algorithms created by the researcher were tested on the TurtleBot robot. The webcam used for the computer vision model is a webcam defined for TurtleBot. In this model designed for mobile robots, the model robot can move without any collision. Results from the ROS-Gazebo simulation environment show that the proposed method is a low cost and flexible method to avoid obstacles.