•  
  •  
 

Abstract

The reliability of Light-Fidelity (LiFi)-based indoor connectivity depends directly on the device orientation and the non-isotropic nature of the optical wireless communication (OWC) channel. In an actual indoor OWC system, the link quality varies significantly with the receiving orientation of the user equipment (UE). During deep-learning-based tracking, this variability imposes considerable computational overhead on resource-constrained IoT microcontrollers when estimating the user's position and orientation. To solve this problem, we employed a discrete Markov Chain estimator to calculate transition probabilities over 3 × 3 m and 5 × 5 m grids. The simulation and experimental results show that the higher the estimation accuracy, the smaller the discrepancy between the estimated and actual angular orientation, and the greater the localization and tracking efficiency. Under identical hardware constraints, comparative results demonstrate that the proposed method outperforms the Random Waypoint (RWP) model, achieving an angular accuracy of 72.4% (RMSE 5.6°), which aligns with the 67–80% simulation range. The performance of conventional deep learning tracking models deteriorates significantly in terms of latency compared to the proposed method. This result highlights that the computational constraints of microcontrollers must be carefully considered during the tracking process, even when high tracking accuracy is achieved. Benchmarking confirms a 2.5–4.8 s latency reduction and a 0.3–0.8 dB SNR gain, enabling real-time field-of-view adaptation.

Keywords

Device orientation, LiFi, Optical wireless communication (OWC), Markov chain, IoT microcontrollers, Real-time tracking, Spatial grids, Field-of-view

Subject Area

Computer Science

Article Type

Article

First Page

2681

Last Page

2699

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Download

Share

 
COinS