駕駛習性學習系統

Driving habit learning system

田智婷

指導教授 ： 方瓊瑤

國立臺灣師範大學/理學院/資訊工程研究所/碩士(2011年)

摘要

本論文研發一套學習各種駕駛者其駕駛習性的系統，主要是因為目前尚未發展出依駕駛習性而量身訂做的駕駛安全輔助系統。現今市面上各車廠推出的駕駛安全輔助系統皆是使用同一標準來評估行車安全，因此本論文希望能透過駕駛習性圖的建立來分析駕駛的特徵，進而為駕駛打造出專屬的安全輔助系統，以符合駕駛的需求也同時維護行車的安全。
　　本系統所建構的駕駛習性圖是由多個行車關係圖組合而成，而行車關係圖包含許多節點和連結，每一個節點代表一個行車因素的大幅改變，每一個連結表示其前後節點的發生順序。從駕駛模擬儀所讀入的原始資料，擷取其行車因素的改變量可形成節點。節點經由階層式合併、權重計算和模糊化的步驟，讓連續的資料轉化為離散的資料表示，進而組成圖像化的行車關係圖。一個行車關係圖代表該駕駛的一個駕駛行為表現，而我們認為行為表現序列能夠表示駕駛的一個駕駛習性，多個駕駛行為表現可以整合成一個駕駛習性圖。而不同的駕駛習性圖可以表示不同類型的駕駛習性。
　　本系統使用交通部運輸研究所的駕駛模擬儀進行實驗。主要因為本研究規劃的實驗具有某種程度的危險性，因此使用駕駛模擬儀實驗可具數據的真實性，也可減低實車實驗對受試人員的風險。實驗結果足以驗證，不同行為表現會產生不同行車關係圖，而不同類型的駕駛也會有不同的駕駛習性圖。上述結論在未來可以給予安全輔助系統更多的客製化空間。

關鍵字

行車關係圖 ； 駕駛習性圖 ； 駕駛習性學習系統 ； 模糊圖像比對 ； 圖像建立

並列摘要

This thesis presents a system to learn the different kinds of driving habits. Recently, no driving assistance system has been considered the characters of the drivers. We develop a driving habit graph which can model the characters of the drivers and hope it can be embedded into the driving assistance system to increase the driving safety.
The driving habit graph is constructed by many driving relational maps which contains nodes and links. Each node represents a significant change of a driving factor, and the links between nodes indicate the ordering of the occurrence time of the nodes. The raw data obtained from the driving simulator are regarded as the driving factors, and a significant change of a driving factor will trigger node creation process which can transform the continuous raw data into the discrete data. A driving relational map constructed by the nodes and links represents a driving behavior. We believe that a sequence of driving behaviors, which can be regarded as a path of the driving habit graph, can model a driving habit. All the paths created from the same driver can integrate to form the driving habit graph of the driver. Different kinds of drivers can construct different types of driving habit graphs.
The experimental data are obtained from a driving simulator, since implementing the experimental scenario on the urban road is dangerous. Moreover, the subjects think that the scenario supplied by the driving simulator is realistic. The experimental results show that different kinds of driving behaviors will correspond to different driving relational maps, and different types of drivers will be represented by different driving habit graphs. In the future, we hope the driving habit graph can make the driving assistance systems more flexible according to the characters of drivers.

並列關鍵字

driving relational map ； driving habit graph ； driving habit learning system ； fuzzy map matching ； graph construction

參考文獻

[Ger 09] B. D. Gerardo and J. Lee, “A framework for discovering relevant patterns using aggregation and intelligent data mining agents in telematics systems,” Journal of Telematics and Informatics, Vol. 26, no. 4, pp. 343-352, November 2009.

[Pau 10] J. Pauwelussen and P. J. Feenstra, “Driver behavior analysis during ACC activation and deactivation in a real traffic environment,” IEEE Transactions on Intelligent Transportation Systems, Vol. 11, no. 2, pp. 329-338, June 2010.

[Cha 09] C. Y. Chang and Y.R. Chou, “Development of fuzzy-based bus rear-end collision warning thresholds using a driving simulator,” IEEE Transactions on Intelligent Transportation Systems, Vol. 10, no. 2, pp. 360-365, June 2009.

[Lin 09] A. Lindgren, A. Angelelli, P.A. Mendoza, and F. Chen, “Driver behaviour when using an integrated advisory warning display for advanced driver assistance systems,”IET on Intelligent Transport Systems, Vol. 3 , no. 4, pp. 390-399, December 2009.

[Kau 08] C. Kaufmann, R. Risser, A. Geven, R. Sefelin, and M. Tscheligi , “LIVES (LenkerInnenInteraktion mit VErkehrstelematischen Systemen) — driver interaction with transport-telematic systems,” IET on Intelligent Transport Systems, Vol. 2, no.4, pp. 294-305, December 2008.