Robust virtual scan for obstacle Detection in urban environments

He Mengwen,Eijiro Takeuchi,Yoshiki Ninomiya,Shinpei Kato

DOI: https://doi.org/10.1109/ivs.2016.7535461

2016-06-01

Abstract:Obstacle detection is an essential technique for intelligent vehicles. Environmental sensing especially plays a vital role to achieve accurate obstacle detection. Unlike classical 2D scan, emerging 3D Light Detection and Ranging (LiDAR) sensors can scan dense point cloud at one time, which represents detailed information of urban environments. The downside of obstacle detection using 3D LiDAR, on the other hand, is its computational cost posed by a large amount of 3D data. The virtual scan (VScan), first introduced by Petrovskaya et al. [1] for efficient vehicle detection and tracking, is a 2D compression of 3D point cloud to represent free space, obstacles and unknown areas. To overcome the computational problem of obstacle detection using 3D LiDAR, therefore, VScan is suitable. In addition, it can bridge across new-born 3D LiDAR sensors and many matured applications based on 2D scan, including occupancy grid map, SLAM, planning, detection, and tracking, due to its 2D representation of 3D point cloud. A key challenge to VScan for intelligent vehicles is that we must improve robustness of VScan in complex urban environments. For example, steep ramps with large slope, low curbs along the road, and overhung barrier gates at the entrance often make VScan mis-behave. In this paper, we present a robust VScan generation method for intelligent vehicles driving in complex urban environments. Our method uses a new data structure, called basic VScan matrix (BVSM), to represent 3D point cloud around the own vehicle. We also develop (i) a simultaneous road filtering and obstacle detection method that works on top of BVSM to generate robust VScan generation, and (ii) a sorted array based acceleration method to perform the VScan generation in real-time. Experimental results from field testing, where steep ramps, barrier gates, and curbs are contained, demonstrate that our method improves the robustness and the speed of VScan generation as compared to traditional methods. The computational cost of generating VScan with 2000 beams is limited within 15ms. The result of obstacle detection on the road with large slopes is stable in the range of 60m with a few number of false positives.