公路客運行車監控之研訂及駕駛與車輛資料庫管理系統之研發—數位式行車紀錄器之應用

Abstract

自數位式行車紀錄器發明以來，已能對行車過程中之駕駛行為作詳細紀錄，並以圖表方式將資訊輸出，惟輸出資訊大多停留在大量報表及數據階段，本研究為使數位式行車紀錄器在駕駛行為管理層面上發揮功用，訂定出燃油消耗類—異常轉速指標、車速不穩指標、衝度異常指標；機件磨損類—冷車啟動指標、煞車異常指標、異常轉速指標；行車安全類—違規超速指標、急加減速指標，車速不穩指標；行車舒適類—前後俯仰指標、車速不穩指標等四大類十一項指標，以實車模擬方式蒐集決定門檻值所需之行車資料，利用模糊德非法的概念，以資料之幾何平均數代表共識值，作為指標門檻值。 本研究在SQL-Server 2000平台上，根據資料庫正規化方法，針對數位式行車紀錄器之原始資料進行第一、第二及第三正規化，利用關聯模型（Entry-Relationship Model）建立各資料表之關聯，設計使用者介面，將本研究所訂定之指標及門檻值之演算方法，轉換為Delphi程式語言，與SQL-Server 2000整合成為一套駕駛與車輛資料庫管理系統，系統包含駕駛基本資料管理模組、異常駕駛監控模組、車輛基本資料管理模組及異常車輛監控模組，各模組間建立關聯，可彼此交叉查詢，亦能針對某一特定指標進行排序，以分析造成異常之主因，並可調閱每位駕駛及每部車之歷史紀錄，以秒為單位進行微觀分析。 最後利用本研究開發之駕駛與車輛管理資料庫系統，對A公司37部裝置數位式行車紀錄器之車輛進行實例分析，取樣時間自2001年11月16日起至2002年4月30日止，經過輔助回歸法檢定指標間相關性後發現，異常轉速指標、車速不穩指標、急加減速指標及違規超速指標等，具顯著之獨立性，接著以統計檢定之方式，分析異常駕駛行為的發生與操作習慣、路段及時間之關係，結果發現異常駕駛行為的發生，並不完全取決於駕駛人的操作習慣，當路況不佳時，對同一位駕駛人，發生異常駕駛行為的機率會相對提高，因此在進行異常監控時，須考慮路況變數造成之影響，才能有效管理駕駛行為。

Not only can the digital vehicle recorders make detailed records on the en route driving behaviors, but they can also present the detail data in charts and tables. Due to lack of practical monitoring indexes and database management system, however, most current digital recorders have presented the outputs in piles of statistical data, which are still of little practical value to a manager. In order to bring the digital vehicle recorders into the practical ground on monitoring and managing the driver’s behaviors, this research attempts to design eleven monitoring indexes which can be divided into four categories: (1) fuel consuming category—abnormal engine rotation, unstable speed gradient, severe jerk; (2) mechanic abrading category—cold engine start, abnormal brakes, abnormal engine rotation; (3) safety category—over speeding, severe acceleration/deceleration, unstable speed gradient; and (4) comfort category—severe jerk, unstable speed gradient. With the fuzzy Delphi method, the index threshold values are determined by the geometric means of the field tests from real bus simulation. On the platform of SQL-Server 2000 with Delphi language, this research employs normalization techniques to transform the raw data of digital recorders into a database. An entry-relationship model is then used to establish the linking among all of the normalized data, monitoring indexes, and threshold values. This database management system contains driver’s basic profile management module, abnormal driver monitoring module, vehicle basic profile management module, and abnormal vehicle monitoring module. Each module is linked with each other so that the users can do the cross query, sort any specific index to analyze the main causes of abnormality, and review the historic records of each driver and vehicle on a per second basis. To validate this driver and vehicle database management system, we choose a bus company for the case study. The en route detail data are automatically collected out of 37 digital recorder equipped buses from 16 November 2001 to 30 April 2002. Auxiliary regression collinearity tests have shown that abnormal engine rotation, unstable speed gradient, severe jerk, and over speeding are statistically independent; while the other seven indexes are highly correlated with these four indexes. Statistical hypothesis tests also conclude that abnormal driver’s behaviors are significantly influenced by drivers as well as traffic conditions. For the same driver, the frequency of abnormality tends to increase significantly in a congested traffic.