- 學位論文
無碰撞條件下自動煞車系統的測試情境設計
Design of Test Scenarios for Automatic Emergency Braking Systems under Non-Collision Conditions
摘要
現今許多科技被用於交通工具上，在汽車上有許多輔助駕駛系統，輔助駕駛系統為了是改善駕駛因閃神或者疲勞造成的錯誤操作，以及提升交通效率與安全性。但電腦計算出的結果總是存在疑點，因此在車輛的生產週期中，測試系統功能也成為重要的一環。現今也存在許多標準來約束系統的設計，讓系統的安全性快速提升，當然也需要評估方法才能夠進行安全性的展示與量化評分，但除了安全性與功能性的評估以外，鮮少舒適性的評分方法，因為舒適性過於主觀，每個人對於舒適性的定義皆不同。 因此本研究將以自動緊急煞車系統(Automatic Emergency Braking System, AEB) 為例，提出一套能夠在無碰撞條件下的測試場景設計，減少測試時的危險性也降低測試的成本，同時定義保守程度的評級的方法。本研究結合MATLAB、Simulink與RoadRunner三方軟體，進行模擬測試的虛擬場景與AEB系統之參數設計，以及評估指標分析，虛擬場景會以現有的AEB測試場景進行重現，並在結果分析完後進行測試場景參數優化。前置作業的部分，除了會介紹測試場景及AEB系統的流程與演算法，還會回顧常見的風險評估方法，找尋系統中具有 缺陷的部分，並找出影響系統表現的重要因子做為測試車輛差異的設定值，並依據AEB演算法之輸入因子與常見的碰撞風險評估指標，作為評估系統保守程度之指標，最後篩選出最適合做為評估AEB系統的測試場景與評估指標組合。除此之外，研究還會針對指標的篩選方法，做最初步的篩選方法建立與介紹，以更快速的找尋理想指標與對應場景。 研究結果顯示，對於AEB系統之保守程度評估，可以採取AEB系統決策指標作為評估參數，以及選擇風險相對較低的場景進行測試。並且部分評估指標會與場景參數有較大的關連性，當然也有指標會因場景之特性無法看出任何差異。而部分車輛系統差異，也會因不直接影響系統決策指標的關係，必須另外找其他適合的場景或者評估指標進行評估。
並列摘要
In modern times, numerous technologies are applied to transportation, with many driver assistance systems integrated into vehicles. These systems aim to mitigate driver errors due to distraction or fatigue and to enhance traffic efficiency and safety. However, there are always uncertainties in computer generated results, making system functionality testing a crucial part of the vehicle production cycle. Numerous standards now regulate system design, rapidly improving safety. Evaluation methods are necessary for demonstrating and quantifying safety and functionality, but comfort evaluation methods are rare due to their subjectivity, as comfort definitions vary widely. This study aims to establish a non-collision test scenario for evaluating Autonomous Emergency Braking (AEB) systems, reducing test risks and costs, and defining a method for rating the system’s conservativeness. The study uses MATLAB, Simulink, and RoadRunner to design virtual test scenarios and AEB system parameters, and to analyze evaluation indicators. Existing AEB test scenarios are recreated for virtual testing, followed by parameter optimization based on the results. Preliminary work includes introducing the test scenarios, AEB system processes, and algorithms, reviewing common risk assessment methods, identifying system deficiencies, and determining key factors affecting system performance as test vehicle differentiation set tings. The study will use AEB algorithm input factors and common collision risk assessment indicators to evaluate system conservativeness, ultimately selecting the most suitable test scenarios and evaluation indicator combinations for AEB systems. Additionally, the study will establish and introduce an initial screening method for indicators, facilitating faster identification of ideal indicators and corresponding scenarios. The findings indicate that AEB system conservativeness can be assessed using decision indicators as evaluation parameters and by testing in relatively low-risk scenarios. Some evaluation indicators show a strong correlation with scenario parameters, while others may not exhibit differences due to scenario characteristics. Vehicle system variations that do not directly impact decision indicators necessitate the selection of other suitable scenarios or evaluation indicators for accurate assessment.