Nonurban Driver Assistance with 2D Tilting Laser Reconstruction

Abstract

In rough environments, such as off-road or post-crisis environments, drivers often need assistance in piloting their vehicles, especially to anticipate obstacles on the driving path. This research aimed to develop such a system, focusing on a cheap and simple method for three-dimensional (3D) reconstruction. This step is important in the detection and classification of negative (under-the-ground) and positive (above-the-ground) obstacles. This information can then be exploited to give feedback to the driver, hence achieving the goal of driver assistance. This article focuses on the 3D reconstruction algorithm, its implementation, and its experimental testing. The choice of sensors is first explained. Because the approach is designed for driver assistance, not for mapping purposes, this leads to real-time and operational constraints (robust and cheap sensors) but eliminates certain other constraints (e.g., dealing with large point clouds). Thus, the selected sensors are a fusion of tilting two-dimensional (2D) LIDAR and stereo cameras. The three-step reconstruction algorithm is then explained. First, the system gets odometry from a stereo pair. Second, 3D points are computed in the ego reference frame with 2D LIDAR scans and servomotor rotation angles. Third, the 3D points are placed in the world reference frame by regularly positioning the previous points on a linear path given by the odometry measurements. The result is a 3D point cloud of the environment in front of the vehicle. Finally, the experimental validation of the approach is explained. A small mobile robot was first tested before applying the approach to a vehicle. Ground-truth acquisitions were conducted to test the veracity of the approach in an outdoor environment. Results show that a coherent map is obtained, but this fusion is not yet suitable for off-road driving at high speeds. However, with some improvements in visual odometry, good 3D reconstruction can be obtained for low- and high-speed driving.