Coded Environments: Data-Driven Indoor Localisation with Reconfigurable Intelligent Surfaces

Indoor localisation in wireless networks poses a unique set of challenges, where traditional approaches often fall short in providing the desired precision and scalability. Our research addresses this challenge by leveraging the capabilities of Reconfigurable Intelligent Surfaces (RIS) in conjunction with machine learning algorithms, specifically gradient-boosted trees (GBT), to achieve remarkable accuracy in indoor localisation tasks.

The experimentation involved deploying directive antennas at a fixed interspacing distance of 1 and 0.5 meters between 9 different positions, with the RIS both activated and deactivated, see Fig.1 and Fig.2. Notably, our results illustrate a substantial enhancement in signal-to-noise ratio (SNR) with the RIS activated, leading to a marked improvement in classification accuracy. For instance, with the RIS activated and an interspacing distance of 1 meter, the average SNR across all positions increased compared to the RIS-deactivated scenario. This increase in SNR directly contributed to an average improvement of 19% in classification accuracy across all positions.

Delving deeper into the classification performance, the confusion matrix presented offers detailed insights into the model's proficiency in distinguishing between positions under varying RIS states. The significant increase in true positives, particularly evident with RIS activation, underscores the pronounced impact of improved SNR on classification accuracy.

Our study illuminates the inherent trade-off between accuracy and the positions’ interspacing distance in localisation frameworks, further underscored by the influence of RIS on SNR. Interestingly, we observed that the impact of RIS activation on SNR varied with the positions' interspacing distance. Specifically, at an interspacing distance of 0.5 meters, the SNR improvement with RIS activation was pronounced, leading to an average accuracy improvement of 17.8%.

Furthermore, our research explores the impact of different parameters, such as subcarrier numbers and channel configurations, on localisation accuracy. Experimental analyses demonstrate the superior performance of the RIS-On scheme across varying subcarrier counts and communication modes. For instance, with an increased number of subcarriers (from 64 to 256), the RIS-On scheme consistently outperformed the RIS-Off scheme, showcasing an average accuracy improvement of 18%.

In conclusion, our study presents a pioneering exploration of RIS-enabled indoor localisation, offering a fusion of advanced technologies to significantly enhance localisation accuracy in next-generation wireless networks. Through comprehensive experimentation and analysis, we unveil the potential of RIS technology in revolutionising indoor localisation, paving the way for future advancements in wireless communication systems. The integration of RIS with machine learning algorithms holds promise for applications requiring precise indoor localisation, such as asset tracking and emergency response systems. Further research is warranted to explore the scalability and real-world deployment of RIS-enabled localisation solutions.


Read the paper in full here:
Coded environments: data-driven indoor localisation with reconfigurable intelligent surfaces | Communications Engineering (nature.com)