3D Point Cloud-Based Virtual Environment for Safe Testing of Robot Control Programs: Measurement Range Expansion through Linking of Multiple Kinect v2 Sensors

Virtual environments are required for safe testing of robot control programs. However, it is important that the virtual environment of the robot simulator adequately reproduces the color and shape information of a real environment. In this study, we use point cloud data obtained from multiple Kinect v2 sensors for building a virtual environment. The coordinate systems of the point clouds obtained from the sensors are unified using the ICP algorithm. Then, a 3D model of the virtual environment consisting of color and shape information of the real environment is generated using Kinect Fusion. Finally, based on the 3D model, the virtual environment is built using the robot simulator Open HRP3. Experiments are conducted to compare the measurement range of the virtual environment that has been reproduced using a single Kinect v2 sensor and four Kinect v2 sensors. Moreover, we calculate the correlation coefficients of the images obtained by a camera sensor in the virtual environment and those obtained by a camera sensor in the real environment. The results show that the measurement range is expanded, and the images of the virtual environment and the real environment exhibit high similarity.