PGPS: A Context-Aware Technique to Perceive Carrier Behavior from GPS Data

This paper presents an approach for perceiving GPS carrier behavior from GPS data only; hence, it is called perceptive GPS (PGPS). The proposed method first extracts behavior feature from GPS data to classify a carrier’s current state. The Newton Hidden Markov Model (NHMM), which integrates Newton’s laws of motion with a Hidden Markov Model, is then introduced to model a GPS carrier’s motion state. On the basis of the NHMM, the PGPS technique records the GPS carrier’s habitual behavior in a Transition Probability Matrix (TPM), which is then used to infer the behavior of the GPS carrier from the online received GPS data. This paper also presents a series of experiments that were conducted to validate the PGPS technique and to determine the proper parameters for the algorithm. By successfully perceiving GPS carrier behavior, the system can provide more friendly services such as improving the accuracy of GPS positioning, providing live view navigation, and detecting if the GPS carriers go astray during navigation.