
Research Article
Attention Based Spatial-Temporal Graph Convolutional Networks for RSU Communication Load Forecasting
@INPROCEEDINGS{10.1007/978-3-030-92635-9_7, author={Hang Zheng and Xu Ding and Yang Wang and Chong Zhao}, title={Attention Based Spatial-Temporal Graph Convolutional Networks for RSU Communication Load Forecasting}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 17th EAI International Conference, CollaborateCom 2021, Virtual Event, October 16-18, 2021, Proceedings, Part I}, proceedings_a={COLLABORATECOM}, year={2022}, month={1}, keywords={VANETs RSU Communication load forecasting ASTGCN Energy schedule}, doi={10.1007/978-3-030-92635-9_7} }
- Hang Zheng
Xu Ding
Yang Wang
Chong Zhao
Year: 2022
Attention Based Spatial-Temporal Graph Convolutional Networks for RSU Communication Load Forecasting
COLLABORATECOM
Springer
DOI: 10.1007/978-3-030-92635-9_7
Abstract
As a special type of base station, Road Side Units (RSU) can be deployed at low cost and effectively alleviate the communication burden of regional Vehicular Ad-hoc Networks (VANETs). However, because of peak hour communication demands in VANETs and limited energy storage, it is necessary for RSU to adjust their participation in communication according to the requirements and allocate energy reasonably to balance the workload. Firstly, tidal traffic flow is generated according to the information of morning and evening peak in the city, so as to simulate the vehicle distribution around RSU on urban roads. Secondly, by inputting the historical information around RSU and the topological relationship between RSU, a network load prediction model is established by using the Attention based Spatial-Temporal Graph Convolutional Networks (ASTGCN) to predict the future communication load around RSU. Finally, according to the forecast of the future communication load, a RSU working mode alteration scheme is proposed with respect to the safety range amongst vehicles in order to control the corresponding area communication load. Compared with other models, our model has better accuracy and performance.