近年來，隨著道路上的車輛越來越多，交通事故發生的數量也逐年增加。 為了有效降低交通事故發生的機率，本篇論文提出了一種演算法，可以用來預測車輛未來軌跡以便及早提醒駕駛碰撞發生的可能性。在車輛路徑預測的研究中，大部分都利用昂貴的外接感測器，但考量成本問題在一般情況比較難以被實作。相反的，我們所提出的演算法會利用簡單又便宜的感測器如GPS或市面上常見的相機來收集資料。 此外我們的演算法也會利用從CAMP所得的車輛迴轉半徑來進行預測，從GPS 所得的經緯度可以用來定位以及預測車輛未來的所在位置，從相機收集到的資料可以用來判斷車輛是否會轉彎。再來我們也會考慮駕駛的反應時間來增進我們預測的準確度。最後實驗結果也顯示利用我們的演算法可以有效地降低預測位置與實際位置的誤差。;As more and more vehicles are on the road, the number of traffic accidents increases each year. To battle with the traffic accidents, one possible mean is to predict the future trajectory of the vehicle so that the driver can be informed when a possible collision is detected in advance. In this thesis, an algorithm for vehicle′s future trajectory prediction is proposed. Although the similar research topic has been studied for many years, most of the existing solutions make use of expensive external sensors, which makes these solutions difficult to be adopted for the general public. On the contrary, our proposed trajectory prediction algorithm collects data from simple and inexpensive sensors like GPS and off-the-shelf camera. The radius of the vehicle from CAMP is adopted in our vehicle path prediction algorithm. The GPS data, such as latitude and longitude, can be used to localize/predict the current/future position of the vehicle. The turn detection data obtained from camera is used to determine if the host vehicle is making a turn. The driver′s reaction time is taken into account as a delay to improve the accuracy of the prediction. The experiment results indicate that the position deviation between the predicted position and the ground truth has been significantly reduced by using our algorithm.
