Online Data Analysis System Design For Driving Behavior Management Of Heavy Trucks

The development of mobile Internet and the popularization of mobile terminals in the era of big data have formed massive mobile object trajectory data.Trajectory data contains abundant spatio-temporal information.Vehicle-mounted data acquisition equipment can not only collect a large amount of vehicle trajectory data,but also obtain OBD data.Thus,the vehicle driving information and the driving behavior of the driver are recorded.The work process the GPS trajectory data,analyze the driving style and driving behavior from the OBD data and design a data analysis system to provide data analysis services for drivers and vehicle managers.Firstly,GPS data is extracted from the vehicle for trajectory data processing.The original trajectory is denoised using median filtering and mean filtering.A map matching algorithm based on hidden markov model is used to further fix the erro of trajectory data.The work uses Douglas-Peuker algorithm to reduce the redundancy of the massive GPS trajectory data so that the amount of the data is reduced by 5 to 10 times while retaining the original trajectory shape characteristics.Stay point detection and trajectory segementation is employed to facilitate data mining.Secondly,the OBD data is extracted to analyze the driving style and driving behavior.The driving styles such as vehicle speed control,engine speed control and steering control are compared using variance and coefficient of variation.Using fuel consumption and other records,combined with additional data such as roads and weather,a driving behavior economic model was established based on the Naive Bayes algorithm.The impact of driving behavior on driving economy was explored,and economic driving recommendations were given;A driving safety model is established for drivers with similar external driving environments.This model punishes different dangerous driving behaviors and calculates the safety factor of the driver on different road sections.Based on the analysis of GPS data and OBD data,this work develops a data analysis system for vehicle management based on big data architecture.The system implements a Hadoop-based vehicle offline data analysis system and a Spark Streaming-based real-time data monitoring system respectively;offline data analysis system uses HBase to store large amounts of data,and performs MapReduce operations on the data regularly to complete data statistics.The data monitoring system analyzes the real-time signal of the vehicle,calculates the real-time driving behavior and fleet statistics.The calculation results are pushed to the frontend for dynamic visualization.Finally,the vehicle off-line data analysis system and vehicle real-time data monitoring system are verified and the key components of the system are optimized.The optimized system is greatly improved on the key indicators before optimization.The test proves that the system can load data processing tasks for 5,000 vehicles.