Impact of video streaming quality on bandwidth in humanoid robot NAO connected to the cloud

This paper investigates the impact of video streaming quality on bandwidth consumption during the transfer of video data from a humanoid robot `NAO' to computing devices, used to perform face recognition tasks, and to the cloud. It presents the results of profiling the network performance of connecting NAO with an edge controller, and discusses the effect of using different qualities of video streaming on the consumed up-link bandwidth. This study considers the limitation of the up-link bandwidth in the Wi-Fi network. It compares the performances of Wi-Fi and Ethernet connections between the NAO robot and a computer. In addition, it examines the accuracy of the face recognition tasks using various streaming scenarios, such as colored video and black & white video. It investigates real-time video streaming using a wide range of frame rates, and video qualities, and their impact on the bandwidth, and accuracy of face identification. The results of our investigations are used to determine the acceptable video quality, frame rate, buffering and bandwidth that would give optimal results in face recognition using NAO robot, and enable efficient data transfer to the cloud.