根據台灣內政部的統計資料，有超過五分之一的死亡車禍是因為駕駛的不注意所導致。在這些死亡車禍中，很多是因為駕駛沒注意前車車況，太晚發現前車因為要轉彎而減速進而導致追撞。為了減少此類交通事故的發生，我們開發了一個轉彎預測系統，本系統能預測即將發生的轉彎事件，以提前通知鄰近車輛，藉此減少交通事故的發生。本系統包括一個車載單元(OBU)及包含數位路線地圖的智慧型手機。我們利用智慧型手機中的感測器來收集車輛的位置訊息，並使用粒子濾波器(Particle filters)來預測車輛未來的位置，然後利用這些位置來計算車輛軌跡的曲率。此外，我們根據位置訊息及數位路線地圖，可以計算出車輛目前所在位置的道路的曲率。藉由比較車輛軌跡曲率和道路曲率的差異以及車輛當時的行駛速度，系統可以判斷車輛是否將要發生轉彎事件。如果受偵測的車輛即將發生轉彎事件，系統便會利用車載單元廣播訊息給鄰近車輛，以警示鄰近車輛避免發生追撞，藉此提升行車安全。實驗結果表示，本系統可以判斷出所有的轉彎事件，且不需要安裝額外的硬體設備，這讓本系統可以容易推廣給大眾。;Many fatal car accidents are rear-end collisions, i.e., drivers were not aware of the car in the front of them making a turn until it was too late. To help alleviate this issue, we develop a lightweight turn prediction system, which identi es the forthcoming turn event at early stage so that the neighboring vehicles can be noti ed in advance to prevent tra c accidents. The system consists of an On-board Unit and a smartphone containing the digital road maps. In the system, the IMU sensors in the smartphone collect position information of the vehicle and predict the future position using particle lters. These positions are then used to compute the curvature of the vehicle trace. In addition, from the position information and digital maps, we can also compute the curvature of the road the vehicle is currently on. By examine the di erence of these two curvatures as well as the speed of the vehicle, the system determines if the vehicle is making a turn at early stage. The experiment results show that the proposed system can identify all the turns perfectly on the road. In addition, the proposed system does not require extra hardware, which makes it possible for inexpensive wide deployment.
