Researchers propose a space-time-aware data recovery network framework

The reporter learned from Hunan University of Science and Technology on April 7 that Liang Wei, a professor at the School of Computer Science and Engineering of the university, joined hands with hunan university, State University of New York and other scientific research teams to successfully propose a space-time perception data recovery network framework. The research results were recently published online in the IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS.

In recent years, Collaborative Intelligent Transportation Systems (C-ITS) have enabled multiple isolated ITS to collaborate to improve the safety, sustainability, efficiency, and comfort of applications. With the expansion of C-ITS systems and wireless communication networks, sensor failures, transmission interruptions, data loss, etc. have become inevitable problems, and will cause serious consequences such as decision-making errors. In this context, data recovery technology is crucial and becomes a prerequisite for many applications.

Data loss patterns are generally divided into three types, including random loss, segment loss, and block loss. In practice, three data loss modes coexist, posing a huge challenge to data recovery technology.

Space-time-aware data recovery network framework. Courtesy of respondents

Performance of individual models in the context of different types of data loss. Courtesy of respondents

In the face of these challenges, the above-mentioned scientific research team proposed a space-time-aware data recovery network framework. The framework includes two parallel feature extraction module graph convolutional layers and output layers, which allow the model to handle spatiotemporal correlations at different scales by overlaying multiple graph convolutions and TCN layers. The network model is suitable for real-time traffic data input scenarios without the need to retrain the entire model. In addition, the network model assembles a graph convolution based on adaptive attention networks of memory and utilizes extended time convolutional networks (TCN) to accelerate training and inference, allowing it to effectively capture spatiotemporal correlations using semantics.

Experimental results show that the framework has good performance in data recovery tasks in C-ITS scenario, and shows lower error than the other four lost data types.

“This research result proposes a traffic data recovery method based on spatio-temporal sensing data recovery network, which can quickly improve the safety and real-time of intelligent transportation application system under different data loss types, and this achievement has an important role in promoting the development of intelligent transportation big data technology.” The reviewer of the paper said.

The research was funded by the National Key R&D Program, the National Natural Science Foundation of China, the Key R&D Program of Hunan Province, and the Natural Science Foundation of Hunan Province. (Source: China Science Daily Wang Haohao)

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button