| 초록 |
There are many cut-in/out and other unexpected situations in the urban setting, so the ability to predict the behavior of nearby objects and to perform algorithms in real time has become important. This is because it is possible to prepare an appropriate real-time response strategy through judgment and control according to the object’s behavior. Therefore, in this paper, two methods for securing real-time stability of self-driving vehicles in the city center are proposed. First, an object tracking method is proposed by using a behavior model-based Kalman filter for location tracking, and a Kalman filter for predicting object speed and behavior. Second, a method for predicting and updating the position of an object at a period of up to 40 hertz by supplementing a lidar sensor with a low period by using the posture data of the vehicle is proposed. This method can be cross-validated through simulation and by using the KITTI benchmark. |
