Research On Fault Detection And Data Repairing For Internet Of Vehicles

In the information age,the science and technology is developing at a high speed as well as the social economy.Because of the deterioration of the city road traffic,the Intelligent Traffic System has a precious chance to develop.As one of the main components of the Internet of Things in the Intelligent Traffic System,the research of the Internet of Vehicles is helpful to improve the traffic situation,transportation efficiency and ensure the safety of commuters.Because the vehicles in the Internet of Vehicles have the characteristics of high-speed mobility and being easily affected by the environment,the data collected by the sensor will inevitably have a variety of problems like modification,lost and forgery.Therefore,the effective fault detection and repairing on the sensor data of the Internet of Vehicles can effectively reflect the real-time state of traffic and then ensure the authenticity of traffic analysis model and the effectiveness of the intelligent traffic management system.On the basis of summarizing the causes of fault data in the Internet of Vehicles,this thesis presents two kinds of algorithms about fault detection and data repairing,which based on dynamic bayesian network model and spatial-temporal correlation combination model.In the scheme based on dynamic bayesian network model,this thesis firstly introduces the structure and parameters learning algorithms of the dynamic bayesian network model,which are the particle swarm optimization algorithm and the expectation maximization algorithm respectively.Secondly,a threshold based fault detection and repairing scheme is proposed according to the data characteristics.In the meanwhile,this thesis also presents a calculation method of the performance analysis about the optimal threshold.Finally,this thesis evaluates the performance of the proposed scheme by simulating the data set that contains the injection fault.The simulation results show that the scheme has low false alarm rate under the condition of high fault detection rate and high correct fault repairing rate.In the scheme based on spatial-temporal correlation combination model,this thesis firstly details the correlation theory and chooses the autocorrelation coefficient and Pearson correlation coefficient as a temporal and spatial correlation index to measure the sensor data respectively.Secondly,this thesis gives a combined model algorithm combining the exponential smoothing algorithm representing temporal correlation and least square estimation algorithm representing spatial correlation to detect the fault data and to repair these data,and provides the method of determining smooth coefficient,weight coefficient and optimal threshold.Then,this thesis makes the combined model algorithm do simulation experiments in the same data set with the original two separate algorithms,and eventually verifying the effectiveness of the proposed algorithm by comparing the simulation results.Finally,the application scenarios of the two schemes proposed in this thesis are given.