High Integrity Lane-level Occupancy Estimation of Road Obstacles Through LiDAR and HD Map Data Fusion

Abstract

In the paper a fast and consistent method to manage uncertainties on detected traffic agents providing reliable results is presented. The information provided by a LiDARbased object detector is combined with a high-definition map to identify the drivable space of the carriageway. Because the use of a HD map requires the use of a localization system, the uncertainty of the estimated pose shall be handled carefully. A novel approach taking into account the localization uncertainty in the perception task by direct propagation of it onto the LiDAR points is proposed. It is compared with a classical propagation that relies on linearized approximation. The good performances of this approach in terms of integrity are demonstrated by the use of real data acquired at the entrance of a roundabout being a particularly complex situation.

Augmented Perception with Cooperative Roadside Vision Systems for Autonomous Driving in Complex Scenarios

“…The detection pipeline is composed of the following main blocks. Further detail about this detection pipeline can be found in [3]: a) Ground segmentation and object clustering: To detect road users from, we first separate the LiDAR point cloud belonging to the ground from the rest. Then, from the remaining points, we exploit a clustering algorithm to individuate and group LiDAR points that belong to a same object.…”

“…To do so, the bounding polygons need to be transformed from the LiDAR sensor frame to the map frame via the AD vehicle localization. The uncertainty of the localization information is taken into account by extending the bounding polygon as in [3].…”

“…Let us consider the LiDAR polygons first. As shown in [3], it is possible to propagate the AD vehicle localization uncertainty directly into the cluster bounding polygons (Figure 1 (c)). The tracker needs a point from the observation to carry on its estimation process.…”

“…Where the function Φ represents a normal distribution. We can then propagate the localization uncertainty of the AD vehicle to point Z k accordingly to [3]. The polygon obtained by the uncertainty injection into Z k is then expressed in the HD map polyline frame in the along and cross directions (see L s and L n in Fig.…”

Spatial Sampling and Integrity in Lane Grid Maps

“…The second one, with uncertainty propagation, a convex hull of each polygon object is augmented taking into account the localization uncertainty. We have used the method presented in [25]. Each vertex of the polygon of an object generates several other vertices with a transformation that takes three times the standard deviation on the pose in each direction.…”