Research On V2V/V2I Communication Assisted Intelligent Transportation Systems

Traffic congestion is an overwhelming problem faced by road travelers all over the world.To make transportation safer and more efficient,projects using advanced technologies such as Intelligent Transportation Systems(ITS)are proposed worldwide,to optimize traffic control systems and alleviate congestion by extracting hidden mobility features of traffic flow.However,the acquisition of vehicle mobility features mostly relies on complex traffic flow models and historical traffic data.In view of the longer sampling time of historical traffic data sets in the existing research,the inability to connect with the real-time traffic,and the traditional centralized signal light control strategy which imposes a heavy computational burden on the Traffic Management System(TMS)server,this paper proposes a Vehicle-to-Everything(V2X)based traffic signal control framework,which extracts traffic data by processing the images on the route planner’s website to find the congested urban hotspots to deploy Mobile Edge Computing(MEC)servers,provides road pheromone information to MEC servers through V2X communication,and accomplishes traffic congestion prediction and intelligent control decision-making of traffic lights to alleviate city-level traffic congestion.The framework is mainly divided into three parts.Firstly,aiming at the problem of IEEE 802.11p channel instability in a highly dynamic environment,we propose an improved Constructed Data Pilots(iCDP)based dynamic channel estimation and equalization scheme to increase the reliability of V2X communication layer.This scheme follows the standard and considers the correlation characteristics between adjacent symbols to construct the data pilots in the time domain and adjacent subcarriers in the frequency domain.The Root Mean Square Error(RMSE)of the proposed iCDP scheme is lower than the Discrete Fourier Transform(DFT),Spectral Temporal Averaging(STA)and the CDP schemes.These results suggest that the proposed iCDP scheme ensures the reliability of communication in a highly dynamic environment.Secondly,aiming at the problem of the real-time detection and prediction accuracy of the city-level traffic congestion in the MEC layer,we devise a deep stacked Long Short-Term Memory(LSTM)network for the multi-point future prediction of congestion in combination with an online learning mechanism.We name the proposed scheme a Fuzzy logic and Deep Learning-based Traffic Congestion Predictor(FDLTCP),to achieve multi-point traffic jam prediction based on city-level real-time traffic images.We compare the proposed predictor with the LSTM,Gated Recurrent Unit(GRU)and Stacked Auto-Encoders(SAEs).Experimental evaluations demonstrate that the proposed FDLTCP has the least Mean Absolute as compared to LSTM,GRU,and SAEs schemes respectively.Thirdly,aiming at the problem of sub-optimal coordination mechanisms and low control efficiency of the existing traffic systems,this study presents the Collaborative Edge Computing-based Traffic Management System(CEC-TMS)in the Traffic Light Control(TLC)layer to address city-wide deployment,and effective collaboration.The proposed algorithm applies multi-agent deep reinforcement learning with a k-nearest-neighbor-based state representation,pheromone-based regional green-wave control mode,and spatial discounted reward to stabilize the learning convergence.We propose PySNS3(a Python-based framework for bidirectional coupling between SUMO and NS3)as a cooperative ITS simulation platform to simulate the deployment of realistic ITS schemes.Finally,we have compared the proposed CEC-TMS with the most recent,Multi-agent Advantage Actor-Critic based TMS(MA2C-TMS),the GREEDY-TMS mode,traditional FIXED,and the Independent Q Learning-based TMS(IQL-TMS)solutions for multiple scenarios,including Monaco and Harbin cities.The proposed CEC-TMS reduces the average waiting for the Monaco city and and the Harbin city(where FDLTCP in the MEC layer indicated the congested hot spots)comparing to the recent MA2C-TMS.The results prove the effectiveness of the proposed framework and deploying MEC servers in the Harbin city.