Research On Data-Driven Real-Time Reliable Urban Vehicular Communications

Vehicular communications,referring to data packets exchange among vehicles and infrastructures,have attracted a lot of attention recently because of its contributions in intelligent transportation systems.To establish ultra-reliable and low-latency inter-vehicle communications is challenging at a city level,according to the features of urban vehicular networks,such as high mobility,dynamic of network topology,giant scales,and frequent disconnections.Existing efforts of inter-vehicle communications are either with assumptions of synthetic road networks,e.g.,Manhattan grids with free moving vehicles,or with a limited scope of traffic pattern analysis over public vehicles,e.g.,taxies and buses.And in past few year,researchers mainly focus on data-driven ve-hicular communications with traffic flow analysis of public vehicles.However,these studies ignored the traffic pattern of urban private vehicles,which are the major enti-ties of urban vehicular networks,leading to a narrow perception for urban inter-vehicle communications.Such methods not only fall short in fully addressing the features of real urban vehicular networks,but also do not meet the requirements of accuracy-and reliability-sensitive applications,e.g.,live HD map in automatic driving.To make up for this blind spot,we propose a data-driven method for urban inter-vehicle communication by analyzing the traffic pattern of urban vehicles.Our work is based on extensive and multiperspective data analysis on real camera surveillance traffic data,which potentially offers a full coverage of various vehicle travel patterns.Based on that,we first propose a novel delay-loss united model,which considers packet loss and estimates packet transmission delays for various types of routes.The proposed delay-loss united model,which selects the path with the least expected transfer delay to transmit the data packets,avoids the loss of the data packets,and achieves real-time and reliable vehicle network communications.Next,we propose another model to evaluate the transmittability of vehicles for data packets transfer with specific destinations and time limits.We use the historical trajectories of urban vehicles to analyze the moving patterns of urban vehicles,and then evaluate the transmittability of vehicles and se-lect the vehicle with the strongest transmittability to transmit data packets.Further,we design an inter-vehicle communication routing system with the integration of the two models which enables efficient and reliable inter-vehicle communications at city levels.Finally,we evaluate our proposals with real citywide datasets of Suzhou Industry Park.We use the trajectories of 840,400 resident vehicles in Suzhou Industrial Park for three months to carry out the simulation experiment of urban vehicle network commu-nications.The results show that our system significantly outperforms state-of-the-art solutions in terms of packet arrival rates and time delays.Compared to GPSR,VADD,TBD and V2VR,our system can improve the absolute value of average arrival rate by 28.32%,29.34%,25.44%,and 19.61%.