Data-Efficient Training for Indoor Positioning Using Light
Author Information
Author(s): Wu Jie, Xu Rui, Huang Runhui, Hong Xuezhi
Primary Institution: South China Academy of Advanced Optoelectronics, South China Normal University
Hypothesis
Can a data-efficient training method improve the performance of Gaussian process regression models for visible light positioning systems?
Conclusion
The proposed Q-AL-GPR method significantly improves positioning accuracy while reducing the required training data.
Supporting Evidence
- The Q-AL-GPR method reduced the mean positioning error from 3.46 cm to 2.76 cm.
- The 97th percentile positioning error decreased from 11.8 cm to 7.5 cm with the proposed method.
- The required number of training data can be reduced by approximately 27.8% using Q-AL-GPR.
Takeaway
This study shows a new way to train models that help find where you are indoors using light, making it faster and more accurate.
Methodology
The study used a three-dimensional visible light positioning system and compared the proposed Q-AL-GPR method with conventional training methods.
Limitations
The study does not address the impact of different LED arrangements on performance.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website