Hybrid map-based SLAM using a Velodyne laser scanner

The first step of environment perception for autonomous vehicles is to estimate the trajectories of the ego vehicle. Based on this, we can build the map of the environment and decide the behaviours of the vehicle in the future. Since a Global Navigation Satellite System (GNSS) is inaccurate and not available in all situations, the Simultaneous Localization and Mapping (SLAM) technique has been introduced to solve this problem. In this paper, we propose a hybrid map-based SLAM using Rao-Blackwellized particle filters (RBPFs). Especially, it was designed with the Velodyne 3D HDL-64 laser scanner which provides rich and accurate data of spatial information around the vehicle. Therefore, our work comprises not only the RBPFs framework but also some signal preprocessing procedures of the sensor. Unlike prior works, we describe the environment by using a grid map and a feature map together rather than using only one of them. Based on both maps, we have formulated a new proposal distribution which is an important performance factor of the algorithm. This makes the uncertainty of a predicted vehicle position decrease drastically and therefore the robustness and efficiency of the algorithm can also be improved. The presented approach was implemented on our experimental vehicle and evaluated in the complex urban scenarios. The test results prove that our approach works well even in real outdoor environments and outperforms traditional approaches.