In this work, the use of Machine Learning methods for robust Received Signal Strength (RSS)-based Visible Light Positioning (VLP) is experimentally evaluated. The performance of Multilayer Perceptron (MLP) models and Gaussian processes (GP) is investigated when using relative RSS input features. The experimental set-up for the RSS-based VLP technology uses light-emitting diodes (LEDs) transmitting intensity modulated light and a single photodiode (PD) as a receiver. The experiments focus on achieving robustness to cope with unknown received signal strength modifications over time. Therefore, several datasets were collected, where per dataset either the LEDs transmitting power is modified or the PD aperture is partly obfuscated by dust particles. Two relative RSS schemes are investigated. The first scheme uses the maximum received light intensity to normalize the received RSS vector, while the second approach obtains RSS ratios by combining all possible unique pairs of received intensities. The Machine Learning (ML) methods are compared to a relative multilateration implementation. It is demonstrated that the adopted MLP and GP models exhibit superior performance and higher robustness when compared to the multilateration strategies. Furthermore, when comparing the investigated ML models, the GP model is proven to be more robust than the MLP for the considered scenarios.

Visible Light Positioning (VLP) is a promising indoor localization technology for providing highly accurate positioning. In this work, a VLP implementation is employed to estimate the position of a vehicle in a room using the Received Signal Strength (RSS) and fixed LED-based light transmitters. Classical VLP approaches use lateration or angulation based on a wireless propagation model to obtain location estimations. However, previous work has shown that machine learning models such as Gaussian processes (GP) achieve better performance and are more robust in general, particularly in presence of non-ideal environmental conditions. As a downside, Machine Learning (ML) models require a large collection of RSS samples, which can be time-consuming to acquire. In this work, a sampling scheme based on active learning (AL) is proposed to automate the vehicle motion and to accelerate the data collection. The scheme is tested on experimental data from a RSS-based VLP setup and compared with different settings to a simple random sampling.

The research interest on indoor Location-Based Services (LBS) has increased during the last years, especially using LED lighting, since they can deal with the dual functionality of lighting and localization with centimetric accuracy. There are several positioning approaches using lateration and angular methods. These methods typically rely on the physical model to deal with the multipath effect, environmental fluctuations, calibration of the optical setup, etc. A recent approach is the use of Machine Learning (ML) techniques. ML techniques provide accurate location estimates based on observed data without requiring the underlying physical model to be described. This work proposes an optical indoor local positioning system based on multiple LEDs and a quadrant photodiode plus an aperture. Different frequencies are used to allow the simultaneous emission of all transmitted signals and their processing at the receiver. For that purpose, two algorithms are developed. First, a triangulation algorithm based on Angle of Arrival (AoA) measurements, which uses the Received Signal Strength (RSS) values from every LED on each quadrant to determine the image points projected from each emitter on the receiver and, then, implements a Least Squares Estimator (LSE) and trigonometric considerations to estimate the receiver's position. Secondly, the performance of a data-driven approach using Gaussian Processes is evaluated. The proposals have been experimentally validated in an area of 3 × 3m$^{2}$ and a height of 1.3 m (distance from transmitters to receiver). The experimental tests achieve p50 and p95 2D absolute errors below 9.38 cm and 21.94 cm for the AoA-based triangulation algorithm, and 3.62 cm and 16.65 cm for the Gaussian Processes.

In this work the design of a transmitter and receiver for indoor visible light positioning is elaborated. The dedicated design is then deployed to conduct measurements to obtain the accuracy of the estimated distance between transmitter and receiver. The results show that the error on the distance estimation between transmitter and receiver is below 3 cm for a realistic setting and good overall performance is achieved. Some additional measurements are done near a reflecting surface to estimate the influence of the none-line of sight component on the received optical power.

Visible Light Communication (VLC) has received substantial research attention in the last decade. The vast majority of VLC focuses on the modulation of the transmitted light intensity. In this work, however, the intensity is kept constant while the polarization direction is deployed as a carrier of information. Demodulation is realized by using a differential receiver pair equipped with mutually orthogonal polarizers. An analytical expression to evaluate the Signal-to-Noise Ratio ($SNR$) as a function of the rotation angle of the receiver is derived. It is demonstrated that the signal quality can deteriorate heavily with receiver orientation when using a single differential receiver pair. A way to overcome this drawback using two receiver pairs is described. The analytical expression is experimentally verified through measurements with two different receiver setups. This work demonstrates the potential of polarization-based modulation in the field of VLC, where receiver rotation robustness has been achieved by means of a dedicated quadrant photodiode receiver.