Research On Characteristics Of Driving Behaviour And Safety Evaluation Based On OBD Information

The driving behaviour of motor vehicle drivers,as the main influencing factor of road traffic safety at present,is a key research target for experts and scholars at home and abroad.For most of the current driving behaviour evaluation models,the selection of evaluation indexes is not comprehensive enough,and the weight of evaluation indexes is difficult to determine,which cannot accurately evaluate the safety of driving behaviour in real time.This paper evaluates driving behaviour through data mining,extracting indicators,constructing a system and using relevant algorithms based on driver manipulation information and vehicle driving conditions information.The main research content is as follows:This study uses On-Board Diagnostic(OBD)equipment as a driving behaviour data collection device.The system collects real-time information on the vehicle status without affecting the driver’s normal driving process,providing a rich and reliable data base for the analysis of safe driving behaviour.Based on the monitored data,the safe driving behaviour evaluation system is constructed,and the human state criterion layer under non-complex working conditions and the vehicle state criterion layer under complex working conditions are investigated separately,and the target layer of the safe driving behaviour evaluation system is evaluated comprehensively by combining the two working conditions.In order to refine the characteristics of driving behaviour according to the types of influencing factors,the distribution patterns of driving behaviour samples are explored in two dimensions,namely driver handling safety and vehicle driving condition safety,in the face of two different working conditions,namely non-complex working conditions and complex working conditions.In the criterion layer of the safe driving behaviour evaluation system,the principles and characteristics of the Fuzzy c-means algorithm(FCM)algorithm are analysed,the Bat Algorithm(BA)is added to optimise the traditional FCM algorithm for its shortcomings,and the traditional BA algorithm is improved,and the improved BA algorithm is tested by a test function.The performance of the BA algorithm is tested by means of a test function,and the clustering effect is compared with the FCM algorithm by means of the Davies-Boulding Index(DBI)to conduct a preliminary study on the evaluation of safe driving behaviour.Qualitative analysis of driving behaviour was carried out to classify the safety risk level of driving behaviour into three levels: safe,risky and dangerous.The driving behaviour evaluation study is carried out using the combination assignment-fuzzy comprehensive evaluation method.Using offline driving behaviour data as samples,the combination assignment method is used to determine the weights of each indicator in the safe driving evaluation system,and the fuzzy comprehensive evaluation method is used to determine the safety risk level of driving behaviour through the feature that the fuzzy comprehensive evaluation method can integrate multiple factors to make a comprehensive evaluation of things.In order to evaluate the safety risk level of driving behaviour in real time and provide a more effective means of evaluating safe driving behaviour,this paper establishes a preclassification-classification model,constructs a pre-classification model using a combined assignment-fuzzy comprehensive evaluation method,and constructs the labelled data for training neural networks.In this paper,a probabilistic neural networks(PNN)is used to classify and identify the safety risk level of the driving behaviour samples and to realise the real-time evaluation of driving behaviour.The Genetic Algorithm(GA)is used to optimize the smoothing factor selection problem in PNN networks,in which the more efficient "binary tournament" selection method is used as the selection operator of the GA.Through experimental comparison,it was verified that the model proposed in this paper can identify the safety risk level of driving behaviour with an accuracy of 99%,and the evaluation time of the model was 0.24 seconds and the evaluation accuracy was 98.67% through real-time data testing.