The Research On Deployment Of Mobile Pollution Source Remote Sensing Network

With the increasing number of mobile pollution sources,the ecological and environmental problems caused by exhaust emissions are increasingly prominent.It is of great significance to obtain the pollutant information in the exhaust in real time to support the urban traffic pollution control and environmental protection,which can be achieved by monitoring the vehicle emissions through the remote sensing devices.These remote sensing devices are widely distributed in the traffic network,which can monitor the exhaust emissions of vehicles on the road in real time.Besides,it does not have any impact on the vehicles during normal driving and prevents many drivers from taking some irregularities to reach the testing standards.However,the practical application of such devices is greatly restricted by the deficiencies of monitoring location strategies.Without doubt,the monitoring performance is influenced by the amount and location of the monitors.Ideally,the more devices we place in the sensing network,the more complete and accurate information we can get.But the cost of the installation and maintenance for the remote sensing devices is rather high,the amount of the devices is thus limited by the economic considerations.Therefore,it requires a trade-off between cost and performance,and allows the devices to be placed optimally.According to different targets of vehicle emission detection,this paper adopts different deployment strategies:(1)In order to detect exhaust emissions from every on-road vehicle,the node deployment problem is mapped to a combinatorial optimization problem,which can be solved by the classical optimization method.(2)In order to monitor the temporal and spatial distribution of vehicle exhaust emissions in the city,the machine learning method is employed.The existing methods based on graph theory transform an optimization location problem into a combinatorial optimization problem,provides a new way to solve the location problem of remote sensing devices.However,the assumption that every vehicle just run on only one road is unrealistic,which means some more roads are actually not imperative to locate monitors.Besides,they do not take the circulation of vehicles inside and outside the sensing field into consideration and it is not sufficient to handle the road network model with unbalance vertexes.In order to solve these problems,this paper proposes a method of location selection of mobile pollution source emission monitoring devices,which considers the boundary nodes of the sensing area.We fully consider the circumstances in which external vehicles enter the monitoring area and the internal vehicles exit the monitoring area,make it capable of measuring the exhaust emissions of vehicles entering the sensing field in real time and restricting the entry of heavily polluting vehicles.It contributes to improve the urban air pollution caused by high-emitting vehicles.Moreover,this approach obtains the result without redundancies and equips the ability to handle the road network model with unbalance nodes.The above method is mainly aimed at the individual exhaust emission of motor vehicles,which cannot achieve the overall monitoring of vehicle exhaust emission in the city.To this end,this paper proposes an algorithm of vehicle exhaust emission monitoring network deployment based on the spatial-temporal graph convolutional network and active learning.Considering the geographic information(road network,interest points,etc.),meteorological data,traffic data,etc.,the proposed algorithm mainly includes two parts: 1)Spatial-temporal graph convolutional network(STGCN)model: In view of the complexity of the traffic network structure,the graph convolution neural network is used to estimate the vehicle exhaust emission of each road section in the sensing area at any time stamp,that is,the exhaust emission estimation of the road section without deployed remote sensing device.2)Active learning(AL)model: add and deploy new monitoring nodes to maximize the accuracy of vehicle exhaust emission prediction.This method takes improving the prediction accuracy of the spatial and temporal distribution of exhaust emissions in the urban area as the basis for the deployment of new remote sensing devices,and realizes the effective prediction of the spatial-temporal distribution of exhaust emissions of motor vehicles on the road section without monitoring device.