Research On Prediction Of Driving Conditions Based On Geographical Features

The vehicle speed conditions directly affect the fuel consumption and emissions of the vehicle.Predicting the long-term speed conditions in the future can optimize HEV energy control strategies and further improve HEV fuel consumption and emission performance.The geographic characteristics of the vehicle have a significant impact on its speed conditions.In order to improve the long-term speed prediction accuracy,it is necessary to consider the effect of the geographical characteristics.Therefore,this paper selects a commuter route in Tianjin as the research object,and explores the influence of geographical features on the vehicle speed conditions,and then explores the prediction method of the long-term speed conditions,which based on the geographical features.The work of the thesis mainly includes the following aspects:(1)Collect condition data and extract feature parameters.Using the vehicle’s GPS information which included in the experimental data and electronic map,select geographical features that have significant impact on the traffic environment of the route.Then,the driving condition fragments of each geographical feature region were extracted,and the appropriate characteristic parameters were selected to analyze the features of the driving condition fragments quantitatively.At last,the characteristic parameters of the driving condition of each geographical feature region were calculated,preparing data for further regular exploration.(2)The influence law of isolated geographical features on the speed of the vehicle was explored.Based on the driving condition fragments of each geographic feature region,the speed state transition matrix was solved.In this paper,the speed state transition rules under different spatial conditions were revealed by using the speed state transition matrix as the mathematical description of speed behavior.Then,the driving conditions were regarded as a Markov chain.Predict the driving conditions of each geographical features with corresponding speed state transition matrix instead of Markov transition matrix,taking the characteristic parameters of the experimental data as error constraints.Based on the predicted driving conditions of each geographical features,a complete route was built.The validity of the complete route was verified and the results show that the average error of the overall characteristic parameters is 4.75%,with 0.24%improvement over the traditional method.(3)In order to improve the applicability of the above method,the influencing laws of similar geographical features on vehicle driving conditions were explored,and then based on the laws,the prediction of the entire driving conditions of the route was realized.This paper established a set of geographical feature vectors that can describe the effect of vehicle behavior.Based on this vector,it was proved that the vehicle driving conditions in the same category of geographic features are similar.Therefore,it is reasonable to cluster the geographic features of the same category.Based on the clusters,the state transition matrixes of each category of geographic features were calculated.At last,use the same method as above to achieve the forecast conditions ofthe entire route with the category state transition matrixes.Verification results of the predicted conditions shows that the average error of the overall characteristic parameters is 5.66%,and the applicability of was the method improved with 0.91%loss of prediction accuracy.