Positioning and Tracking with ODE-LSTM Algorithm for Emerging Smart Rail Systems

Abstract

In this paper, we propose a NLOS positioning and tracking method in order to be applied in the emerging smart rail systems. By analyzing MIMO scatter channels, the geometric relationship for positioning between each UE, BS and scattering point can be modeled that includes information of AOA, AOD, TOA. The accuracy of positioning can be improved by forming an optimization problem with bias in time and orientation. Further, in order to track the mobile UE, an ODE-LSTM algorithm is proposed, which is combined by ODE derivation solver and LSTM network. It puts the positioning information into an ODE-LSTM network to achieve continuous tracking in arbitrary time instance. Simulation Works validate that the proposed ODE-LSTM method shows better performance in nonlinear tracking than traditional Kalman filter or enhanced Kalman filter.