使用蜂巢自動機之車輛移動模型:設計與應用

Abstract

近年來，隨著車輛隨意網路Vehicle Ad-hoc Networks (VANETs)的發展，使得駕駛者能透過VANETs得到很多行車的資訊，進而增加行車安全與交通效率。其中車輛的準確位置、速度與駕駛行為，與VANETs中應用程式習習相關。因此，為了得到準確的車輛移動模式，本論文以蜂巢自動機(Cellular Automaton , CA)車流模式研究混合式車流。過去有學者透過蜂巢自動機研究純汽車、純機車、混合式車流，但僅侷限於直線路段之研究。目前尚無研究使用CA車流模式，研究模擬十字路口的混合式車流之行為。 蜂巢自動機被用於許多領域如交通流模式、物理學、生物學等。在交通流模式多侷限於純小客車流之探討，亦有研究混合式汽機車的模擬，但模擬結果多未符合真實的車流狀況。本論文則是考慮到駕駛者的反應狀況，採用不同實體大小及動態（指機車及小客車）之探討。在本論文中，利用決定每一時間點汽、機車同步縱向前進及橫向位移位置及速率，來模擬汽車與機車在混合車流中的移動行為。為了証明本論文車流模型的可信度，除模擬器之實作外，我們亦實際於現場勘察與蒐集車輛移動資料，並且進行分析。分析結果顯示，使用本論文車流模型之模擬器產生之跟車狀況、換車道狀況、紅綠燈時汽機車行駛的狀況、及轉彎軌跡都符合實際車輛行駛的行為。 研究顯示在VANETs中，若傳輸之車輛與接收之車輛傳輸路徑中有其它車輛阻擋訊號傳輸，路徑之無線訊號能量衰減會大幅增加，因而對傳輸品質會造成很大的影響。於本論文中我們亦發展一套高效能之演算法，於車輛移動模型中計算兩台車之間是否有其它車輛阻擋。此為本論文所提出之車輛移動模型之應用之一。另外，統計資料顯示十字路口是最容易發生碰撞的地方。本論文應用車輛移動模式，另提出一套可預測於十字路口可能產生碰撞的區域及時間之演算法。此演算法可應用於未來的車輛防撞系統上，減少車禍的發生的機率。此為本論文所提出之車輛移動模型之應用之二。

In recent years, with the development of Vehicular Ad Hoc Networks (VANETs) it is possible for the drivers to obtain lots of traffic information while they drive, and in turn both the driving safety and the traffic efficiency are enhanced. It has been shown that the performance of VANET applications is highly correlated to the vehicle’s positions, speed, and driving behaviors. In this thesis, we use the cellular automaton (CA) to accurately model the mixed traffic flows of both cars and scooters. There have been some similar studies in the literature which also utilize the CA to model car-only, scooter-only, or mixed traffic flows on the road; however, in these studies the traffic flows modeled are only for 1-dimensional road segments rather than for the more realistic 2-dimensional road topology with intersections. The CA is utilized in many fields such as vehicle mobility models, physics, and biology. For vehicle mobility models, it has usually been limited to model car-only scenarios. Some do consider mixed scenarios with both cars and scooters; however, the proposed models usually do not capture the nature of the actual traffic flows realistically. In this thesis, to address this problem we take the driver’s reactions into account and differentiate the physical dimensions and the acceleration and deceleration capabilities of cars and scooters. In the proposed model, the movements of cars and scooters in the mixed traffic flow are represented by the X-axis (parallel to the long side of the road) and the Y-axis (perpendicular to the long side of the road) velocities and locations at each time step. To show that our proposed model can generate realistic traffic flows, we also collect real-life traffic flow data from the roads using the LIDAR; comparison of the collected data and the simulator-generated traffic flows shows that car-following and lane-switching behaviors, how the vehicles react to traffic lights, and the turning trajectories of vehicles of both match well with each other. It has been shown that in VANETs, there will be significant additional signal attenuation when the propagation path between the transmitting vehicle and the receiving vehicle is blocked by any other vehicle; the link quality will be significantly lower. In this thesis, we propose two applications to utilize the proposed vehicle traffic mobility model. First, we propose a fast algorithm to calculate if there is any other vehicle blocking the propagation path between two vehicles. Second, statistics show that intersections are the most likely place where the collision can occur. We therefore propose an algorithm to predict the time and area of possible collisions at intersections. The algorithm can be utilized in future vehicle collision avoidance systems to lower the probability of accidents.