| At present,sensors such as videos,radars,inductive loops and microwaves are widely used in freeway accident risk prevention and control system,and these sensors provide basic information for accident risk detection and early warning.However,in practical applications,these sensors are mainly deployed according to engineering experience or with constant spacing,which is subjective and may lead to the lack of sensors on higher-risk road sections or redundant deployment on lower-risk road sections.This not only wastes a large amount of funds,but also makes it difficult to achieve the expected accident risk prevention and control performance.In addition,traffic accidents occur randomly and accidents may happen everywhere on the freeway.However,due to the limited cost,fixed sensors cannot be deployed anywhere on the entire road network.So it is hard to comprehensively obtain freeway accident risk information by using fixed sensors.In recent years,with the rapid development of mobile travel,personal mobile devices such as mobile GPS and vehicular GPS have been widely used,providing a new way for full path accident risk perception.The mobile sensors can compensate for shortcomings of fixed sensors in data collection and are an effective complement to fixed sensors.However,due to the randomness and uncertainty of user demand,travel time,and travel range,the data service quality of mobile sensor is unstable,which makes it difficult to effectively serve accident risk prevention and control systems.Therefore,it is urgent to carry out the study of sensor deployment for freeway accident risk prevention and control,and establish optimal deployment methods of fixed sensors and effective incentive mechanisms of mobile sensors to improve accident risk detection performance and provide support for freeway accident risk prevention and control.Based on this,considering the uncertainty of traffic accidents,this paper proposes an estimation method of accident risk spatial distribution to obtain the spatial distribution of accident risk on freeways and reduce the influence of accident uncertainty,which provides a basis for sensor deployment.To satisfy the requirements of accident scene recognition in practical applications,a sensor deployment method is established for accident scene detection.To satisfy the requirements of indirect inference of accidents,a sensor deployment method is proposed for traffic flow parameter detection.Finally,in order to improve the data service quality of mobile sensors,an incentive mechanism of mobile sensors is designed to improve user participation and enable mobile sensors to effectively serve the freeway accident risk prevention and control system.The main works of this paper are as follows:(1)At present,the differences of road risk are seldom considered in the estimation of accident risk spatial distribution and the limited historical accident samples will induce the biased estimation of risk distribution.To solve the problem,an estimation method of accident risk spatial distribution is proposed based on optimal transmission theory,which improves the accuracy and reliability of spatial distribution estimation of freeway accident risk.Considering the differences in risk characteristics of different types of roads such as tunnels and bridges on freeways,a risk adaptive accident risk spatial distribution function is introduced to obtain the possible spatial distributions of road accident risk.To reduce the bias of accident risk distribution estimation,according to the idea of optimal transmission,an estimation model of accident risk spatial distribution is established by minimizing the difference between the possible spatial distribution of accident risk in every historical period.The international open accident data is used to carry out the experiments,and the proposed method is compared with the fixed bandwidth kernel density estimation method.The results showed that the proposed method is more effective compared with kernel density estimation method.(2)To solve the deployment problem of heterogeneous sensor for accident scene detection under accident uncertainty,a sensor deployment method based on accident scene coverage is proposed.This method could provide a guidance for the deployment of accident scene detection sensors such as video and radar.In this method,to reduce the influence of accident uncertainty,the sensor deployment problem is converted into a coverage problem for accident risk spatial distribution.According to heterogeneous sensor perception model and accident risk spatial distribution,the coverage quality index is established to quantitatively describe the coverage performance for accident risk spatial distribution.An optimal deployment model of heterogeneous sensor is established by considering coverage quality,detection error,deployment cost,and so on.A two-stage solution algorithm based on search pruning strategy and AO-GI-MO is proposed to solve the model.The international open accident data is used to carry out the experiments.The results showed that the deviation of coverage rate on accident risk is very small,which is mainly between 0-10%.Therefore,the proposed method is reliability in practical applications.(3)At present,the temporal-spatial cumulative effect of traffic flow after the accident is seldom considered in sensor deployment,hence the layout of sensors cannot meet the requirement of accident risk detection.To solve the problem,a sensor deployment method is proposed according to the cumulative effects of traffic flow.This method could provide a guidance for the deployment of sensors for traffic flow parameter detection such as inductive loops and microwaves.Firstly,the traffic wave theory is introduced to analyze the evolution process of traffic flow after an accident,and obtain the temporal-spatial cumulative effect of traffic flow after the accident.Based on this,a sensor deployment model is constructed by considering freeway traffic flow,temporal-spatial cumulative effect of traffic flow,temporal-spatial distribution of accident risk,accident detection demand,cost,and so on.And solution algorithm is proposed to solve the model based on improved cuckoo search.The international open accident data is used to carry out the experiments and the proposed method is compared with the uniform spacing method.The results show that the coverage rate on accident of the proposed method is bugger than that of of uniform spacing method,and the improvement is mainly between 0-40%.(4)Currently,the data service quality of mobile sensors is unstable,resulting in the low efficiency in accident risk detection.To solve the problem,a mobile sensor incentive method for accident detection enhancement is proposed.This method could improve the data service quality of mobile sensors and provide reference for incentive issues of mobile sensors in uncertain tasks and user scenarios.Considering the uncertainty of traffic accidents,the incentive problem of mobile sensors is converted into a temporal-spatial coverage problem on accident risk.According to the validity conditions of mobile sensor data,the travel distance and arrival rule of mobile users,and the proportion of mobile sensors in the traffic flow,an opportunity coverage index of mobile sensors on accident risk is established.In order to achieve the expected coverage performance of accident risk,an incentive model for mobile sensors is constructed by minimizing cost with constraint of accident risk coverage rate.The international open accident data is used to carry out the experiments and the proposed method is compared with the uniform incentive method.The result show that the cost effectiveness of the proposed method is improved comparted with the uniform incentive method,and the improvement is mainly between 0-10%.In summary,this study aims to establish the deployment methods of traffic sensors that used in freeway accident risk prevention and control.The proposed method could guide the deployment of sensors in practical applications and improve the safety of freeways.The proposed methods will enrich the deployment theory of sensors in traffic safety prevention and control,and provide reference for the formulation of relevant industry standards. |
PDF Full Text Request