Variational Bayesian EM for SLAM

Designing accurate, robust and cost-effective systems is an important aspect of the research on self-driving vehicles. Radar is a common part of many existing automotive solutions and it is robust to adverse weather and lighting conditions, as such it can play an important role in the design of a self-driving vehicle. In this paper, a radar-based simultaneous localization and mapping (SLAM) algorithm using variational Bayesian expectation maximization (VBEM) is presented. The VBEM translates the inference problem to an optimization one. It provides an efficient and powerful method to estimate the unknown data association variables as well as the map of the environment as perceived by a radar and the unknown trajectory of the vehicle.