Abstract
A highly accurate indoor positioning under the effect of multipath reflections has been a prominent challenge for recent research. This paper proposes a novel indoor visible light communication (VLC) positioning model by connecting k-nearest neighbors (kNN) and random forest (RF) algorithms for reflective environments, namely, kNN-RF. In this fingerprint-based model, we first adopt kNN as a powerful solution to expand the number of input features for RF. Next, the importance rate of these features is ranked and the least effective one(s) may be removed to reduce the computation effort. Next, the training process using the RF algorithm is conducted. Finally, the estimation process is utilized to discover the final estimated position. Our simulation results show that this new approach improved the positioning accuracy, making it nearly five times better than other popular kNN algorithms.
Highlights
In the last few decades, the global positioning system (GPS) has been widely employed in positioning and navigation because of its high reliability and accuracy, and because of its real-time positioning capability [1]
In the random forest (RF) algorithm we can improve the performance of the training process and reduce the computation cost using the feature importance ranking [21,22]
visible light communication (VLC) positioning model is proposed in this paper, namely k-nearest neighbors (kNN)-RF
Summary
In the last few decades, the global positioning system (GPS) has been widely employed in positioning and navigation because of its high reliability and accuracy, and because of its real-time positioning capability [1]. The main contributions of this paper are as follows: Most of recent VLC-based research have ignored the impact of reflection (see Section 2). This noise exerts a very bad influence on the positioning accuracy, especially in areas outside the room’s center. We propose a novel machine learning-based indoor positioning solution This method first uses kNN as a powerful tool to expand the number of features for RF in the step. The position of the mobile object can be located based on the estimation process in the RF algorithm which collects all the features directly from both the LED lights and the kNN output signals.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
