Performance Limits for Fingerprinting-Based Indoor Optical Communication Positioning Systems Exploiting Multipath Reflections

Abstract

Multipath reflection degrades the performance of visible light communications (VLC) based localization systems, where it is often considered as a strong random noise. However, due to the inherent transmission features of light, the optical wireless indoor channel is static; therefore, multipath components can be modeled as deterministic functions of the transceiver location, furnishings, and room geometry. In this paper, we investigate the performance limits of fingerprinting-based localization with multipath reflection as a source of information, i.e., a fingerprinting map. Limits on the localization accuracy are determined using the Cramer-Rao lower bound (CRLB) for different numbers of photodetectors deployed in the system and received signal features captured. The tightness of the analytical CRLB is tested by comparing it to the performance of a fingerprint-based positioning algorithm that uses the nearest neighbor method. Simulation results show an achievable root mean squared positioning accuracy of 45 cm and 5 cm (for one and four photodetectors, respectively), for an empty room. We then investigate the practical limitations on localization accuracy caused by a narrow transceiver bandwidth. Numerical results show that the localization system can still achieve decimeter accuracy for system bandwidths of 200 MHz, which makes fingerprinting schemes practical for off-the-shelf infrared devices.

Cooperative Passive Pedestrian Detection and Localization Using a Visible Light Communication Access Network

“…It also experiences a more deterministic and stable behavior of the optical wireless channel, leading to superior accuracy compared with RFbased counterpart techniques. Theoretical studies [13]- [15] and experimental demonstrations [5]- [7] verify that visible light-based localization systems outperform RF-based systems by orders of magnitude.…”

Optical Channel Impulse Response-Based Localization Using An Artificial Neural Network

“…in which 𝜌 is the PD responsivity, the energy per pulse is E p , 𝑅 p = 1/𝑇 p is the pulse repetition rate, where 𝑇 p is the pulse repetition period, 𝐺(𝑡) the Gaussian-shaped pulse defined in [2], ℎ (𝑞) PD the low-pass filter that models the 𝑞th PD, and ℎ LED is the LED impulse response. ( * ) denotes the convolution operation.…”

“…In (2), 𝑛 (𝑞) (𝑡) is additive zero-mean Gaussian noise, where the factor √︁ 𝑁 p accounts for the 𝑁 p independent and identically distributed (i.i.d) zero-mean terms in the pulse-train. In an optical wireless environment, ambient light is the predominant noise source, adding shot noise with variance calculated based on [2,Eq. 4].…”

“…VLC localization has also attracted substantial attention due to the broad use of light-emitting diodes (LED) for indoor lighting. Optical channel impulse response (OCIR)-based VLC positioning methods benefit from the largely static and deterministic optical wireless channel for localization [1][2][3][4][5]. This paper uses an artificial neural network (ANN) to evaluate the effectiveness of employing multipath reflection information within the OCIR with varying levels of temporal resolution for localization.…”

“…However, these techniques do not take advantage of multipath reflection patterns to reach the ultimate localization accuracy. In our initial research studies [2,3], we presented a proof of concept and fundamental limits for OCIR-based localization using a simple nearest neighbor algorithm for localization. In [3], we proposed an infrared uplink positioning algorithm to identify the user location information from the multipath reflections as captured by the fingerprinting map of a small selection of features of the OCIR.…”