Temporal-Graph-BA Sed Routing Algorithm In Software Defined Vehicular Networks

With the rapid breakthrough of the automobile industries and telecommunication technologies,the Intelligent Transportation System（ITS）and the Vehicular Ad-Hoc Networks（VANETs）gain much more attention.However,the traditional distributed architecture slow down the evolution of VANET due to its limited vision.Thus a brand-new network paradigmSoftware-Defined Vehicular Networks（SDVNs）improve the communication performance of the vehicular networks greatly.Whereas,most existing routing schemes in SDVNs consider the networks as a sequence of static graphs.Routing in the static graph could be inefficient and inaccurate.The main reason could be the temporal information that would play a vital role in the vehicular network.Meanwhile,the evolution of the internal structure of the network presents a regular pattern with time flows,which can be discovered under the SDVN architecture.Thus,in this paper,we propose three temporal-graph-based routing algorithm in SDVNs.They are similar to each other in terms of algorithm architecture.They consider the vehicular network as a temporal graph where every data transmission as an edge has its specific temporal information.The first is the Temporal-information-based Adaptive Routing Algorithm（TibAR）.First,distance and historical routing information are used to construct a Markov Model（MM）.in order to represent and predict the future vehicular network.After that,a two-way prediction algorithm is utilized to efficiently and accurately predict the possible future routing,and the future routing temporal graph can be constructed.Finally,the temporal graph shortest path algorithm is used to obtain the optimal routing path.For better performance,the prediction model and the corresponding prediction algorithm are innovated.First,the Hidden Markov Model（HMM）is used instead of MM,the intersection is set as the hidden state,the vehicles around the intersection are set to the observation.Dynamic planning and greedy strategies are used to construct the HMM that represents the current vehicle network.Then we use a new prediction algorithm:computing the optimal and suboptimal intersection paths respectively,fully connecting and updating the temporal information of qualified vehicles around these intersections to construct a future routing temporal graph.Finally,the temporal graph shortest path algorithm is used to compute the optimal routing path too.Finally,in the third routing algorithm:Novel GCN-based Greedy Routing Algorithm（NGGRA）,the routing architecture is first innovated.The controller trains a decision model based on the information collected from the data plane,and distribute it to the data plane.The vehicle computes and decides the routing path by itself through this model.The decision model incorporates the parameter of node importance into the traditional Graph Convolutional Network（GCN）,and proposes a nodeimportance-based Graph Convolutional Network（NiGCN）.Finally,a distributed greedy routing algorithm based on this network is proposed.The simulations results show that,compared with other state-of-the-art work,the three routing algorithms have greatly improved computing efficiency,delivery rate,transmission delay and routing quality.