Action Recognition System Based 5G Communication and Edge Computing for Service Robot

Abstract

Action recognition is one of the most critical technologies for the robot to provide service excellently. In order to solve the problems of low accuracy, limited arithmetic power, and low intelligence of service robot's action recognition, an action recognition system based on 5G communication and edge computing is proposed. Communication and data interaction between robot and edge computing device is realized by using 5G routers and technologies such as NAT and Socket. A LightSlowfast network is designed for the service robot, which improves action recognition accuracy while reducing the computational effort. The experimental results show that the Light-Slowfast network achieves 72.9% accuracy on the HMDB51 dataset. The system performs well in practical tests, and the low latency of data transmission can satisfy the robot's demand for real-time action recognition.