Towards Real-time Radio-SLAM via Optimal Importance Sampling

Abstract

In future cellular networks, it will be possible to estimate the channel parameters of non-line-of-sight propagation paths providing unique opportunities for simultaneous localization and mapping (SLAM) with commodity user equipments (UEs). Radio-SLAM generally relies on generating samples of the UE trajectory and constructing a trajectory-conditioned map. To reduce the number of samples and complexity, we propose an iterative method to approximate the optimal sampling density. The numerical results demonstrate that the added computational complexity of the proposed method can be easily justified by the more efficient use of particles. As an outcome, the presented filter nearly achieves the lower bound and still runs in real-time.