Research On Safety Prediction Model Of Visual Secondary Task Driving Based On BP Neural Network

Secondary task driving is one of the main reasons for occupant's attention and driving distraction.Prediction of driving safety for secondary task driving is very important for reducing traffic accident rates.Secondary tasks are divided into three categories: visual,cognitive and psychological.In the driving process,as the driver obtain road information mainly by the visual way,therefore,the visual tasks effect more on driving safety than cognitive or psychological tasks.In order to achieve the safety prediction of secondary task driving rapidly and effectively,this paper takes three kinds of visual driving task as the research object.The "passenger car driver behavior and traffic safety research platform" of Jilin University,Institute of Transportation,were utilized to carry out the experiment.With eye movement parameters set and vehicle driving state parameter set for the safety evaluation index,were obtained by the platform.Then,the safety grade evaluation method of the main task driving and three secondary task driving,under the intercity freeway scene,were investigated.Based on these,the safety prediction model of secondary task driving was established and the applicability test was passed.The main contents of this paper are as follows:In order to obtain the safety evaluation level of driver's secondary task driving.It is necessary to establish a scientific and concise characteristic evaluation index system.According to the actual situation,with the help of the experimental platform,we designed the corresponding experimental scheme.Then 40 drivers' eye motion parameters and the vehicle running state parameters by road simulation experiment were collected.The safety evaluation index formed by the data processing parameters after data processing is analyzed,and the changing rules of the parameters under different driving tasks were analyzed.Application of grey clustering theory,combined with the formation of a clustering algorithm,instead of equivalence relation in rough set theory were achieved.The algorithm has complementary advantages to make dimensionality reduction to optimize the evaluation indexes.Then the characteristic evaluation index system was established.Based on the characteristic evaluation index system.F-AHP and middle type membership function were utilized,respectively,for calculating the weights of characteristic evaluation indexes and the single factor fuzzy judgment matrix,of each driver.Then take the bounded operators for the algorithm,the calculated safety evaluation of digital level was obtained.In order to establish the safety prediction model of secondary task driving,the internal principle and BP neural network learning rules were analyzed in this paper.Then the appropriate network parameters were established.The data of characteristic evaluation indexes was used for the input sample of BP neural network,while the evaluation digital levels of driver's safety was used as neural network output training sample of 30 drivers.Debugged the network parameters continuously until error requirement was permitted.Finally,the remaining 10 drivers' data were applied to verify the applicability of the safety prediction model.The results showed that the visual secondary task driving safety prediction model established in this paper has good applicability.