Online Matrix Factorization-Based Traffic Flow Prediction Empowered by Edge Computing for the CAVs

The Connected and Autonomous Vehicles (CAVs) is considered to be a promising technology to improve the traffic congestion. However, to realize its expected benefits, the real-time, accurate and forward-looking guidance message are required. Based on this, considering various practical constraints and the characteristics of the CAVs fully, multiple distributed edge computing servers are deployed at the network edge to provide the real-time storage and computation support in our paper. Furtherly, based on the fruitful deployment for edge computing servers, an efficient traffic flow online prediction model is constructed, which can adaptively modify according to the dynamic changes of the actual road traffic state, thus providing more accurate and forward-looking results. The simulation results based on MATLAB platform show the proposed deployment scheme for edge computing servers only needs more cost than the enumeration method. Moreover, compared with other baseline methods, the online prediction model proposed can improve 37.5%–76.9% in terms of prediction accuracy.