In this paper, we propose an obstacle avoidance algorithm for autonomous vehicles traveling along the global path that safely avoids its front obstacles and return to the global path. With the point cloud data obtained using the LiDAR sensor while driving, points corresponding to specific angles in the front LiDAR channels are extracted, and the distance between the points of each channel and the mathematically calculated value are compared to determine the presence of obstacles. The target steering angle is determined, at which there is no obstacle for the car to drive through, is used to control the car to perform the avoiding operation. The steering angle data based on incorrectly recognized result may cause a sudden steering operation. To prevent this sudden disturbance, the final target steering angle is smoothed using a Kalman filter. However, if only the above method is used, autonomous vehicles may invade sidewalks or cross the centerline. Therefore, in this paper, we precisely measured autonomous vehicle's position and heading using Extended Kalman Filter and High Definition Map to prevent the vehicle from invading undesired area by activating or deactivating certain range of angles to maintain the vehicle within the drivable area.

Keywords: local path planning, high definition Map, LiDAR sensing, positioning