GAN-based RSSI Correction for Visible Optical Links

The received signal strength indication (RSSI) is a crucial metric in both communication and positioning within visible light communication (VLC) systems. Current RSSI-based VLC systems are limited in their ability to correct errors through received optical power and distance fitting, and require recalibration after photodiode (PD) conversion. In response to this problem, this paper proposes an RSSI correction method based on the generative adversarial network (GAN). This method transforms channel modeling into a parameter distribution learning problem, thereby providing a key performance index for estimating Lambertian parameters, calculating optical transmit and received optical power, and characterizing channel attributes through the quantification of received signal strength. Experimental results show that the proposed method achieves an NMSE as low as 0.0029 in ranging applications and a positioning accuracy of $\mathbf{1 1. 2 3 ~ c m}$ when applied to visible light positioning systems.