Accuracy-Oriented Data Offloading for Cooperative Video Analytics in Internet of Connected Vehicles

Onboard cameras are widely deployed in vehicles for enhancing their sensing abilities and further supporting their real-time control on the road. However, the sensing range of a single vehicle is limited and even blocked by the inevitable occlusion among the moving vehicles. Cooperative computing is...

Full description

Saved in:
Bibliographic Details
Published in2024 IEEE/CIC International Conference on Communications in China (ICCC) pp. 1263 - 1268
Main Authors Huang, Jianhua, Rong, Shichu, Zhong, Weifeng, Kang, Jiawen, Huang, Xumin, Yu, Peng, Liu, Yuan
Format Conference Proceeding
LanguageEnglish
Published IEEE 07.08.2024
Subjects
Accuracy
Connected vehicles
Cooperative computing
data offloading
Road side unit
Roads
video analytics
Visual analytics
Wireless communication
Wireless sensor networks
Online AccessGet full text

Cover

More Information
Summary:Onboard cameras are widely deployed in vehicles for enhancing their sensing abilities and further supporting their real-time control on the road. However, the sensing range of a single vehicle is limited and even blocked by the inevitable occlusion among the moving vehicles. Cooperative computing is an effective and promising solution for this single-vehicle issue. This paper studies cooperative video analytics for a cluster of wireless connected vehicles. First, the raw video frames generated on the vehicles are downsampled to reduce the data size, improving the efficiency of communication and computing. The downsampled video data are then processed by using the collective computing resources of the vehicles and road side units (RSUs). A nonconvex optimization problem maximizing the overall video analytics accuracy in the vehicle cluster is formulated subject to the latency constraints. To quickly solve the problem, an efficient three-phase auction-based method is proposed, making decisions on optimal data downsampling and offloading. Results demonstrate the time-efficiency of the proposed method and its advantage in boosting video analytics accuracy by efficiently exploiting the limited bandwidth and computing resources in the vehicle cluster.
DOI:10.1109/ICCC62479.2024.10681779