In the thesis, we addressed the main issues of robot movement in the vicinity of humans - the ability to detect people and consequently the safe movement of the robot for all participants in the area. We tested an open-source algorithm developed under the European project Spencer and evaluated its effectiveness in a real environment with appropriate sensor equipment. We focused most of our attention on environment detection with an RGB-D camera, defining its field of view and evaluating the accuracy of person detection.
After downloading our ROS package, we installed the appropriate drivers for the RGB-D camera on the computer and the necessary supplementary code (specifically our launch file, which opens the Rviz environment, displaying the camera data in real-time). In the next step, we recorded the data captured by the RGB-D camera and saved it (in a rosbag file). The obtained data was then analyzed and presented in the Matlab environment. We varied the height of the camera and repeated the process.
The analysis was successful and allowed us to determine the area with practically 100\% reliability of operation and partial reliability.
|
