Environment recognition based on multi-sensor fusion for autonomous driving vehicles

Unmanned driving of an autonomous vehicle requires high reliability and excellent recognition performance of the road environment and driving situation. Since a single sensor cannot recognize various driving conditions precisely, a recognition system using only a single sensor is not suitable for autonomous driving due to the uncertainty of recognition. In this study, we have developed an autonomous vehicle using sensor fusion with radar, LIDAR and vision data that are coordinate-corrected by GPS and IMU. Deep learning and sensor fusion improves the recognition rate of stationary objects in the driving environment such as lanes, signs, and crosswalks, and accurately recognizes dynamic objects such as vehicles and pedestrians. Using a real road test, the unmanned autonomous driving technology developed in this research was verified to meet the reliability and stability requirements of the NHTSA level 3 autonomous standard. © ICROS 2019.