Design and analyze one object recognition system with human in the loop for unmanned platforms with 5G communication
Xueyin Fang, Xuesong Wu, Yue Qian · 2021
Unmanned platforms (unmanned aerial vehicles, unmanned vehicles, etc.) have been widely used in various fields. They are generally controlled by human-in-the-loop remote control systems. The video returned by them will greatly help people's work and life, but traditional communications bring challenges to high-speed, ultra-low-latency unmanned platform applications. The large-bandwidth, high-reliability, and low-latency communications provided by 5G technology can well meet the application requirements of human-machine cooperation in unmanned platforms. Visual search represented by object recognition is a very common task in daily life. Although current automatic algorithms can complete part of the task, they have weak understanding and poor environmental adaptability; human operators have strong comprehensive perception capabilities and good environmental adaptability. However, the accuracy is low, individual differences are large, and fatigue is easy. Based on 5G technology, achieving human-machine cooperation can significantly improve system performance. This paper models the process of the human-in-loop multi-object recognition task in unmanned systems and designs four different ways of collaboration between human and expert systems. In addition, we propose to switch the four collaboration manners can be switched dynamically. Especially, based on human factors engineering experiments, this paper analyzes the improvement of the overall system performance with the help of the expert system and the impact of different collaboration manners on the efficiency in the target recognition task. They can provide support for the collaboration between human and unmanned platforms in the context of 5G applications.