An HMM-Based Localization Scheme Using Adaptive Forward Algorithm For LTE Networks

User location is important for network operators to perform the network management and services. This paper considers the problem of fingerprinting localization in the LTE networks. Fingerprint information is extracted from measurement reports (MRs) gathered at the network side. We present an Adaptive Forward (AFW) localization algorithm based on a hidden Markov model (HMM), where the hidden states are the locations of the User Equipments (UEs), and the observations are the Reference Signal Received Power (RSRP) vectors extracted from MRs. The transition probability of the HMM is trained based on the neighboring relationship between the RSRP vectors to eliminate the need for any sensor or additional information. Furthermore, in AFW algorithm, the adaptive switching strategy between forward algorithm and single point localization is proposed to improve the localization accuracy by suppressing the potential accumulated errors. Extensive experiments operated with data sets from real LTE networks show that, the median localization error of the proposed algorithm is around 23m for outdoor UEs, which is comparable to the related works which need additional information.