建立二維駕駛反應模型以評估市區路口風險

網路與雲端技術加快了自駕車發展的速度，許多趨勢均表示電腦將逐漸取代人類駕駛。在逐步取代的過程中，勢必經歷交通環境中電腦與人類駕駛的共存期，若無法評估自駕車進入交通環境的影響，將會影響人類駕駛的用路權也會傷害到自駕車的競爭力。本研究將行駛行為與影響評估的重點放在「風險」，而環境與用路所形成的風險與駕駛因應的風險行為決定了交通的樣貌，因此本研究提供一模擬方法，建立模擬路口環境與駕駛的互動行為，並分析風險趨勢。本研究使用指數分佈函數計算二維空間下的機率風險，並以刺激反應模型做為駕駛反應模型，並設計不同的駕駛積極度，以模擬實際交通環境中不同種類的駕駛。完成指定交通環境的模擬及設定後便可針對區域進行風險分析。本論文探討之風險分析包含三個層面:提供駕駛完成決策的駕駛風險評估;針對不同行駛路線所累積之路線風險評估;從道路環境的角度評估各種車輛，在環境中交互影響所形成的環境風險評估。駕駛風險評估可觀察駕駛風險來源，並歸納衝突種類;路線風險評估可觀察路線中，風險變化的趨勢，並分析高風險事件的聚集區域;環境風險評估則可觀察整體環境中風險分佈狀態。本論文的二維風險模型亦可評估交通政策改良前後的風險評估結果，我們以路口禁止左轉為例，觀察執行後的駕駛、路線以及環境的變化，並得知禁止左轉後高風險區域風險降低，且風險分佈較為平緩。此模擬方法不僅限於路口風險模擬，在地圖有路線幾何的條件下亦可應用於圓環等其他型態的交通環境，也能夠為自駕車提供量化的風險指標，作為行駛策略的參考指標。

關鍵字

二維機率風險；駕駛反應模型；駕駛積極度；駕駛風險評估；道路風險評估；路口風險評估

並列摘要

Autonomous vehicle technology has accelerated rapidly due to the ad- vance of network and cloud computing. Trends have shown that computers are gradually replacing human drivers. During the transition, human drivers and autonmous vehicles have to coexist on roads. Evaluating the impacts of autonmous vehicles in traffic with human drivers becomes critical to maintain the right of human drivers as well as the competitive of autonmous vehicles. This thesis models risks that caused by driver behaviors when thay interact, along with the environment that decides the quality of road transportation. The exponential distribution funciton is used to calculate probability of risk in two-dimensional space, while the stimulus-response model is used as driver reaction model. We design different aggressiveness to characterize various drivers. Risk analysis in this thesis can be divided into three aspects, the driver risk evaluates the risks from drivers’ prospective, the road risk evaluates the risk associated with different routes, the environment risk evaluates the risks distribution within a region of interist. The driver risk can provide not only the risk level a driver might experience, but also the direction of risk. The road risk can provide as a desision to which route provides the minimal risk. The environment risk provides an overall evaluation show the whole risk dis- tribution in the environment. This thesis also investigates the results of a policy change in road sections. We compare the addition of no left-turns in road sections as a policy change, and found out that the value of high-risk regions’ decreases, and the distribution of risk are more flat then it was. We belive our work can be applied not only in road sections, it can also be use in roundabout or other kind of road environments, given their well- defined geometry. The evaluation of risk on autonmous vehicles can therefore be achieved with different driving strategies.

並列關鍵字

two-dimensional probability risk ； driver reaction model ； driver aggressiveness ； driver risk evaluation ； road risk evaluation ； road seciton evaluation

參考文獻

[1] A. Anand, “Top self driving car companies in 2022.” =https:// www.analyticssteps.com/blogs/top-self-driving-car-companies-2022, 2022.

Google Scholar

[2] SAE international, “SAE J3016 automated-driving graphic.” =https://www.sae.org/ blog/sae-j3016-update, 2021. [Online; accessed 10-Jan-2022].

Google Scholar

[3] R. Hussain and S. Zeadally, “Autonomous cars: Research results, issues, and future challenges,” IEEE Communications Surveys Tutorials, vol. 21, no. 2, pp. 1275–1313, 2019.

Google Scholar

[4] M. Bertoncello and D. Wee, “Ten ways autonomous driving could redefine the au- tomotive world,” McKinsey Company, vol. 6, 2015.

Google Scholar

[5] W. Official, “Waymo safety report.” = https://waymo.com/intl/zh-tw/safety/„ 2022. [Online; accessed 28-Jan-2022].

Google Scholar

主題瀏覽