New method improves indoor location accuracy using light

A transfer learning approach helps devices find their position more efficiently in factories.

Researchers developed a method for indoor localization using Visible Light Communication (VLC). This approach uses transfer learning (TL) to improve how devices find their location.

The method uses transfer learning (TL) to handle environmental variability. This includes physical obstacles and lighting fluctuations. The researchers also introduced a model efficiency (ME) metric. This metric combines localization accuracy, energy efficiency, data efficiency, and transfer gain.

Conventional localization techniques struggle in dynamic environments. These methods require frequent data collection and environment-specific calibration. Such methods can limit scalability and interoperability. They also impose significant energy penalties.

Testing on a real-world dataset from a BOSCH factory showed results. The model achieved a 47% improvement in localization accuracy compared to a conventional model without TL. The TL models achieved up to 67.5% ME. Conventional models achieved only 10.4% ME under high-noise conditions.

This approach may deliver a highly efficient and scalable indoor localization system. It is suitable for energy-constrained industrial applications. The expansion of the Internet of Things makes indoor localization a critical component for industrial automation.

Key facts

  • The model achieved a 47% improvement in localization accuracy over conventional models.
  • TL models reached up to 67.5% ME compared to 10.4% for conventional models in high-noise conditions.
  • The method uses transfer learning (TL) to address lighting fluctuations and physical obstacles.

All sections

Edition