Limits...
An Indoor Mobile Location Estimator in Mixed Line of Sight/Non-Line of Sight Environments Using Replacement Modified Hidden Markov Models and an Interacting Multiple Model.

Ru J, Wu C, Jia Z, Yang Y, Zhang Y, Hu N - Sensors (Basel) (2015)

Bottom Line: Consequently, to improve precision, each moving node uses the IMM model to integrate the results from the HMM and its modified forms.Simulation experiments conducted show that our proposed algorithms perform well in both distance estimation and coordinate calculation, with increasing accuracy of localization of the proposed algorithms in the order M-HMM, RM-HMM, and HMM + IMM.The simulations also show that the three algorithms are accurate, stable, and robust.

View Article: PubMed Central - PubMed

Affiliation: School of Information, Northeastern University, Shenyang 110819, China. rujingyu@hotmail.com.

ABSTRACT
Localization as a technique to solve the complex and challenging problems besetting line-of-sight (LOS) and non-line-of-sight (NLOS) transmissions has recently attracted considerable attention in the wireless sensor network field. This paper proposes a strategy for eliminating NLOS localization errors during calculation of the location of mobile terminals (MTs) in unfamiliar indoor environments. In order to improve the hidden Markov model (HMM), we propose two modified algorithms, namely, modified HMM (M-HMM) and replacement modified HMM (RM-HMM). Further, a hybrid localization algorithm that combines HMM with an interacting multiple model (IMM) is proposed to represent the velocity of mobile nodes. This velocity model is divided into a high-speed and a low-speed model, which means the nodes move at different speeds following the same mobility pattern. Each moving node continually switches its state based on its probability. Consequently, to improve precision, each moving node uses the IMM model to integrate the results from the HMM and its modified forms. Simulation experiments conducted show that our proposed algorithms perform well in both distance estimation and coordinate calculation, with increasing accuracy of localization of the proposed algorithms in the order M-HMM, RM-HMM, and HMM + IMM. The simulations also show that the three algorithms are accurate, stable, and robust.

Show MeSH

Related in: MedlinePlus

The two models proposed in this paper.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4507710&req=5

sensors-15-14298-f009: The two models proposed in this paper.

Mentions: The MTs in this paper refer to robots or human, and their velocities usually range from 0 to 4 m/s. We thus divide the velocities into low- and high-velocity parts in order to simplify our model. The low-velocity part is a Gaussian function with mean value zero and variance 1.5, and the high-velocity part is one with mean value three and variance 1.5. Note that the narrower the spaces into which we divide the velocity, the more precise the results we obtain. However, our two-part strategy is effective in describing the trajectory of the terminal in this narrow range of velocities. Figure 9 shows the Markov switching model, which shows that the system varies between the low- and high-velocity models, where Pab represents the Markov transition probability from mode a to mode b.


An Indoor Mobile Location Estimator in Mixed Line of Sight/Non-Line of Sight Environments Using Replacement Modified Hidden Markov Models and an Interacting Multiple Model.

Ru J, Wu C, Jia Z, Yang Y, Zhang Y, Hu N - Sensors (Basel) (2015)

The two models proposed in this paper.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4507710&req=5

sensors-15-14298-f009: The two models proposed in this paper.
Mentions: The MTs in this paper refer to robots or human, and their velocities usually range from 0 to 4 m/s. We thus divide the velocities into low- and high-velocity parts in order to simplify our model. The low-velocity part is a Gaussian function with mean value zero and variance 1.5, and the high-velocity part is one with mean value three and variance 1.5. Note that the narrower the spaces into which we divide the velocity, the more precise the results we obtain. However, our two-part strategy is effective in describing the trajectory of the terminal in this narrow range of velocities. Figure 9 shows the Markov switching model, which shows that the system varies between the low- and high-velocity models, where Pab represents the Markov transition probability from mode a to mode b.

Bottom Line: Consequently, to improve precision, each moving node uses the IMM model to integrate the results from the HMM and its modified forms.Simulation experiments conducted show that our proposed algorithms perform well in both distance estimation and coordinate calculation, with increasing accuracy of localization of the proposed algorithms in the order M-HMM, RM-HMM, and HMM + IMM.The simulations also show that the three algorithms are accurate, stable, and robust.

View Article: PubMed Central - PubMed

Affiliation: School of Information, Northeastern University, Shenyang 110819, China. rujingyu@hotmail.com.

ABSTRACT
Localization as a technique to solve the complex and challenging problems besetting line-of-sight (LOS) and non-line-of-sight (NLOS) transmissions has recently attracted considerable attention in the wireless sensor network field. This paper proposes a strategy for eliminating NLOS localization errors during calculation of the location of mobile terminals (MTs) in unfamiliar indoor environments. In order to improve the hidden Markov model (HMM), we propose two modified algorithms, namely, modified HMM (M-HMM) and replacement modified HMM (RM-HMM). Further, a hybrid localization algorithm that combines HMM with an interacting multiple model (IMM) is proposed to represent the velocity of mobile nodes. This velocity model is divided into a high-speed and a low-speed model, which means the nodes move at different speeds following the same mobility pattern. Each moving node continually switches its state based on its probability. Consequently, to improve precision, each moving node uses the IMM model to integrate the results from the HMM and its modified forms. Simulation experiments conducted show that our proposed algorithms perform well in both distance estimation and coordinate calculation, with increasing accuracy of localization of the proposed algorithms in the order M-HMM, RM-HMM, and HMM + IMM. The simulations also show that the three algorithms are accurate, stable, and robust.

Show MeSH
Related in: MedlinePlus