A LiDAR-based real-time capable 3D Perception System for Automated Driving in Urban Domains

Abstract

We present a LiDAR-based and real-time capable 3D perception system for automated driving in urban domains. The hierarchical system design is able to model stationary and movable parts of the environment simultaneously and under real-time conditions. Our approach extends the state of the art by innovative in-detail enhancements for perceiving road users and drivable corridors even in case of nonflat ground surfaces and overhanging or protruding elements. We describe a runtime-efficient pointcloud processing pipeline, consisting of adaptive ground surface estimation, 3D clustering and motion classification stages.Based on the pipeline's output, the stationary environment is represented in a multi-feature mapping and fusion approach. Movable elements are represented in an object tracking system capable of using multiple reference points to account for viewpoint changes. We further enhance the tracking system by explicit consideration of occlusion and ambiguity cases.Our system is evaluated using a subset of the TUBS Road User Dataset. We enhance common performance metrics by considering application-driven aspects of real-world traffic scenarios. The perception system shows impressive results and is able to cope with the addressed scenarios while still preserving real-time capability.

Enhanced Perception for Autonomous Driving Using Semantic and Geometric Data Fusion

“…Typically, these use high density LiDARs for gathering both 3D information as well as some texture information. In [10], the authors present a traditional solution based on a single high-density LiDAR, tailored for urban environments, that separates stationary features (represented in a multi-layer grid) from the dynamic objects which are subsequently tracked. The authors of [11] also argue for the utility of multi-layer grid maps when presenting a method for semantic segmentation of LiDAR scans using a convolutional neural network.…”

“…The back-projection technique can be less appropriate when the camera and LiDARs have significantly different mounting positions/viewpoints on the ego vehicle, and requires additional time for the projection and image space fusion. In [10] obstacle measurements from a single high density LiDAR are clustered using a polar grid with vertical compression (each cell can contain multiple occupied intervals). Voxel based representations are also proposed to represent and process LiDAR measurements, but mainly for static/dynamic discrimination and with single high-density LiDAR setups.…”

Obstacle detection and tracking algorithm based on multi‐lidar fusion in urban environment

“…But it is easy to detect remote obstacle as multiple targets. Rieken et al [18] designed a real-time 3D LiDAR perception system for autonomous driving based on adaptive ground surface estimation, 3D clustering and motion classification stages.…”

“…Step 1: For the unassociated obstacle from the previous frame, its historical tracking rate R track is firstly calculated by Equation (18).…”