Research For Stand-out Driving Behavior Modeling And Good Parameters Discovery Method Based On Bus Database

Driving behavior optimization not only lowers occurrence probability of traffic accident, but also improves energy efficiency level and reduces pollutant emissions. Current studies indicate that corresponding data processing method is core issue to realize driving behavior estimation. Vehicle intelligent terminal collects vehicle driving status data through a series of vehicle sensor and data acquisition equipment and processes them to generate driving behavior evaluating results, while a related suggestion is given to remind driver.Unfortunately, due to the randomness of driving, reasonable data processing model is difficult to establish. In this paper, fixed route and same model buses are selected as modeling object. According to the characteristic of bus driving data, a new concept, driving key region, is put forward. Then, based on driving setting key region concept, a good driving behavior parameters discovery method, which aim to provide real-time operation recommendations in this type of key regions, optimize driving technology, reduce fuel consumption and improve safety and comfort, is proposed. The main contents of the thesis are as follows:First of all, current research status of automatic driving, assistance systems and driving behavior analysis modeling are introduced.After that, the driving behavior model is established and the details of bus energy-saving principles are studied. Moreover, fuzzy comprehensive evaluation is employed to obtain driving behavior assessment output.Then, framework for good driving behavior is constructed, while the data collection system of bus is introduced in detail. According to the prior rule and evaluation criteria, the relative good driving data is extracted. Furthermore, based on the concept of key region, extracted data of different time period is divided into several slices, which is disposed by MapReduce parallel framework. Both clustering and pruning processing are used to deal with data of each slice to get good driving behavior parameters.Finally, a driving data set, which not include in training data set is selected to validate the model and parameters. The validation results show that the model and parameters discovery method are effective. An application scenario is introduced, and the test results of different time periods are given out.