Mapping Extended Landmarks for Radar SLAM

In the early years, algorithms Simultaneous localization and mapping (SLAM) using automotive developed for landmark-based SLAM operated under the assumption radar sensors can provide enhanced sensing capabilities that each landmark can generate at most one measurement for autonomous systems. In SLAM applications, with per scan (point landmark), and therefore mainly focused a greater requirement for the environment map, information on estimating the location of landmarks. However, with on the extent of landmarks is vital for precise navigation and the advent of high resolution automotive radar sensors, landmarks path planning. Although object extent estimation has been may occupy multiple radar resolution cells and hence successfully applied in target tracking, its adaption to SLAM generating more than one radar detection per scan, i.e. an remains unaddressed due to the additional uncertainty of the extended landmark. By exploiting information such as the location sensor platform, bias in the odometer reading, as well as the of each radar detection, as well as their spacial spread, measurement non-linearity. In this paper, we propose to incorporate we can also estimate the size and orientation of a landmark, the Bayesian random matrix approach to estimate in addition to its centroid location.