Autonomous Navigation in Rural Environment based on Digital Filters and LIDAR Data

Road detection is one of the essential problems in autonomous driving. Common approaches for road detection are heavily based on predetermined highly accurate digital maps. Such approaches work good in urban locations. Outside of them, in rural environment, building of such maps as well as their storing and transmission are very challenging tasks. Moreover, rural environment is known to vary rapidly due to natural and human factors. These factors limit the implementation of the state-of-the-art approaches suitable for urban environments. In this paper, we present the approach which performs road surface detection based on sensor perception system without detailed prior maps. We detect road borders on the obtained surface by computing the distance from a car to all surrounding points and process this data by digital filters and observation of distance dashes. After road detection calculation of car position relative to road borders is performed. Such approach allows generating feasible trajectories for the vehicles to reach path waypoints. The generated trajectories are updated taking into account vehicle odometry and IMU data. We demonstrate the performance of the approach on a full-scale autonomous vehicle navigating in rural environment. Obtained results proved the approach allows the vehicle to navigate in rural environment reliably and at a high speed.