Multipath Channel Estimation and Pattern Recognition for Environment-Based Adaptive Tracking

The ability of a GNSS receiver to adaptively change its tracking strategy based on the surrounding environment will enable robust positioning in wireless and mobile applications under challenging signal conditions. Analysis of the received Line-of Sight (LOS) signal and multipath signals indicated that their characteristics depend on the surrounding environment. This paper introduces several algorithms that utilize this dependency to enable adaptive tracking. A Markov Chain Monte Carlo (MCMC) based algorithm is introduced to estimate the multipath channel. A machine-learning pattern-recognition algorithm is introduced to identify the surrounding environment from the multipath channel characteristics. A tracking strategy selector module is then activated to decide on a tracking strategy that best suits the current signal condition. The channel impulse response is modeled by a Bernoulli-Gaussian process to accommodate sparse multipath signals, where the presence or absence of a multipath signal at a delay is modeled by a Bernoulli random variable. The MCMC-based algorithm estimates the channel impulse response, which is used to construct a multipath pattern that can be analyzed. The analysis is done using a multi-class pattern recognition algorithm, which performs two main functions: feature extraction, which is done using a Principal Component Analysis (PCA) approach; and pattern classification, which is done using a multi-layer feed-forward Probabilistic Neural Network (PNN) approach. The tracking strategy selector module utilizes both closed loop tracking and open loop tracking to accommodate various patterns. The open loop tracking is activated in unstable or rapidly changing signal conditions. The algorithms are verified using real and simulated GPS signals with accurate multipath models.