Social Prediction-Based Handover in Collaborative-Edge-Computing-Enabled Vehicular Networks
Lei Guo, Qingyang Song, Shupeng Wang, Zhe Liu, Lei Guo
Abstract
Collaborative edge computing (CEC) can realize the cooperation and integration of heterogeneous resources distributed in adjacent areas, increasing the overall resource utilization efficiency. In a CEC-supported heterogeneous vehicular network composed of different access solutions, including cellular vehicle-to-everything (C-V2X) and dedicated short-range communications (DSRC), good network connections can guarantee timely access to edge resources. How to maintain stable and high-quality network connections for vehicles is a crucial issue. With traditional received signal strength (RSS)-based handover schemes, vehicles may encounter severe ping-pong effects and even direct handover failures leading to data packet loss. In this article, to overcome the frequent handover problem caused by vehicles’ high-speed motion and the ever-changing network environment, we propose a trajectory prediction-based handover scheme. In this scheme, the sojourn time of a vehicle staying in each candidate network’s coverage can be obtained through a social long short-term memory (social-LSTM)-based prediction model. Together with the signal strength, available bandwidth, and cost, the sojourn time is also taken as a handover decision attribute parameter. Simulation results show that our proposed scheme can reduce the number of handovers effectively.