Autonomous Vehicle Research
To safely navigate autonomous vehicles, it is essential to have precise terrain map with further enough detection range for undertaking appropriate control actions. The terrain map should also contains accurate enough position data of the obstacles to secure the movement of the vehicle even in narrow passages. Our lab has been developing algorithms to build a static and dynamic map of the driving environments including lane and road markers by using Lidar such as HDL-42E velodyne, SICK LMS series and Ibeo Lux.
As the image processing technology gets more developed, the importance of cameras on board is getting increased. Our lab has developed Image-processing technology, which can detect traffic signs and lights, lane stripes, and other objects.
Path planning is a high level module in autonomous vehicle and a component that affects directly vehicle’s motion. For safe autonomous driving, it is necessary to develop a path planning algorithm that can generate a proper driving path. A variety of path planning algorithms for the motion control of autonomous vehicles have been proposed and applied in the past. Our autonomous vehicle also have applied various path planning algorithms such as A*, Potential field method, RRTs and etc.
Localization is the most important technology of autonomous vehicle to recognize the position where it is. Generally GPS is used for localization system in outdoor environments. But to get accurate position data from GPS, high cost GPS device is needed. For developing the accurate localization system with low cost GPS device, our lab has developed GPS/INS system by using GPS, odometry and IMU data. Our lab also has developed not only GPS-based localization system, but also localization system with perception system such as SLAM, AMCL and etc.
Vehicle Platform and Control
Our lab has been developing robust autonomous vehicle platform for testing developed perception, localization and planning algorithms. Our vehicles are based on SUVs from Hyundai/Kia Motors. A custom made interfaces enables direct electronic actuation of all the driving interfaces. Vehicle data are sensed automatically and communicated to the computer system through a ethernet or CAN bus. Our lab also have developed motion controller which allows our vehicle to robustly follows the reference trajectory.