Optical wireless 3D-positioning and device orientation estimation

Accurate sensing and localisation are considered as necessary features of future communication systems, including 6G. To harness the full potential of radio frequency (RF) and optical wireless communication (OWC), the localisation of user devices is essential, which further facilitates efficient beam steering, handover, and resource allocation. In this paper, we have considered a practical scenario where users are mobile with random device orientation. A convolutional neural network (CNN) is introduced to estimate the user position and orientation based on the received signal strength (RSS). CNN demonstrates superior performance in optical wireless positioning by proficiently extracting features from only RSS data. According to the simulation results it is observed that, by adjusting the structure of the dataset, a significant improvement in the estimation of the location is obtained in comparison with previous methods. We also consider having the noisy orientation data from the device sensors and investigate localisation performance in such a scenario. Finally, the impact of configuration of access points (APs) on the model is studied. This work demonstrates that a low-complexity accurate localisation, with average error as low as 1.8 cm, is indeed feasible.