An optimal self-learning estimator for predicting inter-cell user trajectory in wireless radio networks

Prediction of the next-crossing cell is an important issue for mobility and connection management in wireless radio networks. We propose a new approach to modeling inter-cell user mobility, and develop an optimum self-learning estimator for trajectory tracking and next-crossing cell prediction. The dynamic states of movement, in terms of speed and position, are obtained by modeling the user´s acceleration as a time correlated, semi-Markovian process, and by passing subsequent signal-strength measurements to neighboring base stations through an extended, self-learning Kalman filter. Prediction of next-crossing cell is obtained by evaluating user dynamic states with cell geometry. Analysis and simulation results show that our prediction algorithm is robust in the presence of pass loss, shadow fading, and random movement, being able to predict the position, speed, and direction-of-travel of the mobile user with a high degree accuracy