Using lighting design tool to simplify the visible light positioning plan and reduce the deep learning loading.

Area of Science:

• Computer Science
• Electrical Engineering
• Optical Engineering

Background:

• Indoor positioning systems (IPS) are crucial for various applications.
• Visible Light Positioning (VLP) offers a promising alternative to traditional IPS.
• Data acquisition for training VLP models can be time-consuming and resource-intensive.

Purpose of the Study:

• To develop a method for reducing the training data requirements for indoor Visible Light Positioning (VLP) systems.
• To adapt existing commercial lighting design software into a VLP design tool.
• To evaluate the effectiveness of deep learning models trained with simulated data against those trained with experimental data.

Main Methods:

• Commercial lighting design software was modified into an indoor VLP design tool.
• Deep learning models, including second-order regression, fully-connected neural networks (FC-NN), and convolutional neural networks (CNN), were employed.
• VLP models were trained using both experimentally acquired datasets and datasets generated by the modified software.

Main Results:

• The adapted software effectively generated datasets for VLP model training.
• Similar positioning accuracy was achieved when VLP models were trained with software-generated data compared to experimental data.
• The proposed method significantly reduces the effort required for collecting training datasets for indoor VLP.

Conclusions:

• The transformation of lighting design software into a VLP design tool is feasible and effective.
• Using software-generated data for training VLP models substantially lowers the burden of experimental data collection.
• This approach facilitates the development and deployment of indoor VLP systems by simplifying the training process.