本研究的目的為建立一套方法來針對各種不同車型的車輛進行多維度的動態模型參數辨識。由於在車輛模型中，有些參數的數值非常難以得到，或是需要相當先進且昂貴的設備才能夠進行量測。所以，本研究建立一套方法來獲得這些參數的數值，此一方法包含一實車測試流程並針對所擷取的動態變數來進行數據的分析以得到所擬辨識的參數。所進行的數據分析主要是利用最佳化的數據分析方法，以致於以所辨識出來的參數代入動態模型所進行之模擬與實車測試所量測得到的動態變數相同。
本研究針對擬辨識的參數依據其會影響的動態來規劃實車測試，所規劃的測試包含三類，分為縱向測試、驅使垂向運動之測試以及橫向測試。然後，再利用所擷取蒐集的相關動態變數進行參數的辨識。於參數的辨識的過程當中首先設定各個參數的辨識初始值以及誤差範圍，然後利用各種實車測試所擷取之動態變數做為目標，以最佳化的方法進行數據的分析來縮小各個參數的可能落點範圍至目標範圍 之內，最後，再利用一個受到所有參數值影響的實車測試動態變數來進行最終的參數辨識。
最終的結果顯示，以本研究所辨識出之參數數值代入動態模型進行時速120公里、測試長度200公尺的雙車道變換模擬所擷取之Pitch Rate、Roll Rate、Yaw Rate及側向加速度動態變數，與實車測試所得到的動態變數作比較。結果顯示，Pitch Rate平均每一筆擷取數據之相對誤差為-0.236%，其餘則為Roll Rate的0.608%，Yaw Rate的-0.225%，以及側向加速度的0.200%。

The purpose of this study is to establish a parameter identification method for various high dimensional vehicles dynamics model. In the dynamics model some parameters values are often not directly available from measuring instruments. Therefore, this study establishes a method to infer these parameters values. This method used the dynamic response of the actual vehicle test to parameter identification by data analysis. The data analysis is based on optimization, so that the simulation of the dynamic model with the identified parameters is the same as the dynamic response by the actual vehicle test.
This study planned tests based on the parameters that affect the dynamic to be variables. Which contain longitudinal tests, driving vertical motion tests, and lateral testing three category. Then use the dynamic variables collected of the tests to identify the parameters. We set up the initial value and the error range of each parameter. And use dynamic variables collected of the actual vehicle test as a goal for narrow the error range down to . Finally, the parameter identification is performed using a dynamic variables collected of the actual vehicle test which is affected by all parameter.
We use the pitch rate, roll rate, yaw rate, and lateral acceleration dynamic variables measure by Double Lane Change which of 120 km/hr and 200 meters length to validate. And the dynamic variables of the parameters identified by this study are compared with the dynamic variables by the actual vehicle test. The final results show that the average relative error of pitch rate is -0.236% each records, roll rate is 0.608% each records, yaw rate is -0.225% each records and, lateral acceleration is 0.200% each records.

摘要 ...................................................I
Abstract .............................................III
謝誌 ...................................................V
目錄 ..................................................VI
圖目錄 ................................................IX
表目錄 ...............................................XIV
第1章 緒論 ..............................................1
1.1 研究背景與動機 ......................................1
1.2 文獻回顧 ............................................1
1.2.1 建立車輛動態模型之相關文獻 .........................1
1.2.2 參數辨識相關文獻 ..................................6
1.2.3 本研究使用之實驗計畫法簡介 .........................9
1.2.4 本研究使用之最佳化算法簡介 ........................11
1.2.5 混合最佳化算法 ...................................17
1.3 研究目的 ...........................................18
第2章 研究方法 .........................................19
2.1 擬辨識之模型規劃 ...................................20
2.2 擬辨識之目標參數規劃 ................................21
2.3 目標參數初始值及可能落點範圍設定 .....................23
2.4 實車測試規劃 .......................................25
2.4.1 縱向測試 .........................................27
2.4.2 驅使垂向運動之測試 ...............................30
2.4.3 橫向測試 .........................................31
2.5 目標參數可能落點範圍優化 ............................34
2.6 最終參數辨識 .......................................42
第3章 模擬結果 .........................................45
3.1 目標參數初始值及可能落點範圍設定 .....................45
3.2 Carsim模擬實車測試之動態變數結果 ....................46
3.2.1 縱向測試所擷取之動態變數結果 ......................46
3.2.2 驅使垂向運動之測試所擷取之動態變數結果 .............49
3.2.3 橫向測試所擷取之動態變數結果 ......................51
3.3 目標參數可能落點範圍優化之結果 .......................57
3.4 最終參數辨識之結果 .................................71
3.5 動態模擬誤差 .......................................75
第4章 結論 .............................................79
參考文獻 ...............................................80
附錄一 目標參數初始值設定 ...............................82
附錄二 參數辨識結果之誤差 ...............................83
作者簡介 ...............................................84

[1]Ni, J., and Hu, J., 2017, “Dynamics control of autonomous vehicle at driving limits and experiment on an autonomous formula racing car,” Mechanical Systems and Signal Processing, Vol. 90, pp. 154–174.
[2]Wu, L., Ma, F., Pu, Y., and Yin, H., 2017, “Integrated Effects of Active Suspension and Rear-Wheel Steering Control Systems on Vehicle Lateral Stability, ” SAE Technical Paper, doi:10.4271/2017-01-0257.
[3]許齡元，2011，車輛動態估測與預測系統，博士論文，國立交通大學，機械工程學系，新竹。
[4]Ryu, J., Gerdes, J. C., 2004, “Estimation of vehicle roll and road bank angle”, American Control Conference, pp. 2110-2115.
[5]蕭期元，2012，主動式前輪轉向控制系統之模式建立與控制設計，碩士論文，國立中山大學，機械與機電工程學系，高雄。
[6]Hakan, Gökdağ, 2014, “Vehicle Parameter Identification Using DE Algorithms,” Journal of Multidisciplinary Engineering Science and Technology, Vol. 1, Issue. 5, pp. 225-229.
[7]Thaller, S., Reiterer, F., Schmied, R., Waschl, H., Kokal, H., and del Re, L., 2016, “Fast Determination of Vehicle Suspension Parameters via Continuous Time System Identification,” IFAC-PapersOnLine, Vol. 49, iss. 11, pp. 448–453.
[8]Zhang, Y., Wang, L., Zhang, B., and Zhang, N., 2015, “Vehicle Parameter Estimation Based on Full-car Dynamic Testing,” SAE Int. J. Passeng. Cars - Mech. Syst, doi:10.4271/2015-01-0636.
[9]Castro, A. A., Braga, S. L., and Neto, M. S., 2013, “Estimation of Dynamic Parameters of Ground Vehicles based on Genetic Algorithms,” SAE Technical Paper, No. 2011-36-0636.
[10]Cavazzuti, M., 2013, “Design of Experiments,” in Optimization Methods, pp. 11-42, Springer-Verlag Berlin Heidelberg, Germany, Berlin.
[11]施才憲，2001，企業應收帳款敏感性分析之研究，碩士論文，國立中山大學，財務管理學系研究所，高雄。
[12]邢文训，謝金星，1999，現代優化計算方法，清華大學出版社，北京，第1-95頁。
[13]Storn, R., and Price, K., 1997, “Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces,” Journal of Global Optimization, Vol. 11, pp. 341–359.
[14]李維平，江長育，蔡宛庭，2011，「搭配擾動策略之差分演化演算法」，資訊科技國際期刊 (IJAIT)，第5卷，第1期，第24-39頁。
[15]Kirkpatrick, S., Gelatt, C. D., Vecchi, M. P., 1983, “Optimization by Simulated Annealing,” Science, New Series, Vol. 220, No. 4598, pp. 671-680.
[16]李世炳，鄒忠毅，2002，「簡介導引模擬退火法及其應用」，物理雙月刊，第廿四卷，第二期，第307-319頁。
[17]Mori, B., de Castro, H., and Cavalca, K., 2004, “Application of Simulated Annealing Algorithm in the Reliability Optimization of Redundant System, ” SAE Technical Paper, No. 2004-01-3280.