近年來，隨著交通事故發生的數量日漸增加，如何及早預防交通事故的發生漸漸變成一個相當重要的議題。本論文提出了一套車輛路徑預測的演算法，此演算法計算出車輛的未來軌跡後，並利用預測的軌跡來偵測車輛在未來一段時間內是否有可能會與其他車輛發生碰撞。我們所提出的方法是利用高精準度的定位系統來蒐集車輛資訊。由於蒐集的部分資料會有數值起伏較大的情況，所以我們利用三種不同的平滑資料方法來處理資料並增加預測結果的準確性。在實驗中，我們蒐集三個資料集來評估演算法的效能。實驗結果也顯示此演算法有效地降低預測的車輛位置與實際的車輛位置之間的誤差距離。;As the number of traffic accidents increases in recent years, how to prevent traffic accidents becomes more and more important. In this thesis, we propose a vehicle trajectory prediction algorithm which can be used to detect future collisions in advance. The proposed algorithm uses the high precision GPS built in the ITRI WAVE/DSRC Communications Onboard Unit to collect the vehicle information. Because part of the collected data appear to fluctuate, three different smoothing procedures are used to process the data to improve the accuracy of the prediction result. In the experiments, three datasets are collected to evaluate the performance of the prediction algorithm. The experiment results show that the proposed algorithm significantly reduces the deviation between the predicted position of vehicle and the ground truth.
