AI-based strategies for handover enhancement in visible light communication systems
Nenhuma Miniatura disponível
Data
2025-03-21
Autores
Camporez, Higor Araújo Fim
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal do Espírito Santo
Resumo
The Internet of Things (IoT) growth, particularly applications involving wireless devices, has significantly increased the demand for signal bandwidth. However, Radio Frequency (RF) wire less systems presented a limited spectrum to support massive device connections and suscepti bility to electromagnetic interference. These challenges have increased the interest in exploring alternative solutions to face RF issues while maintaining high data rates, low latency, reliability, and cost efficiency. Advancements in Light Emitting Diode (LED) technology have introduced highly energy-efficient lighting capable of high-speed modulation of light intensity. Thus, these characteristics have driven research into Visible Light Communication (VLC), which can uti lize existing lighting infrastructures for data transmission using a broad and unregulated optical spectrum (≈ 400 THz). Additionally, VLC can also provide physical layer security, low power consumption, high transmission speeds, and immunity to RF electromagnetic interference. Spectral efficiency and high data rates are critical for VLC systems, with Orthogonal Fre quency Division Multiplexing (OFDM) emerging as a robust and spectrally efficient modulation technique for indoor applications. However, nonlinearities introduced by multicarrier signals in LED-based systems can degrade performance. To address these issues, techniques such as Constant-Envelope OFDM (CE-OFDM) have been developed to mitigate Peak-to-Average Power Ratio (PAPR), improving power efficiency and reducing distortions, particularly in high power transmission scenarios. Additionally, VLC faces several challenges, including signal blockage by opaque objects, confinement of signals, and limited Access Point (AP) coverage. Addressing these limitations often requires deploying ultra-dense networks to ensure reliable connectivity across large areas. However, such dense deployments can lead to frequent han dovers, increasing infrastructure costs and complexity. This thesis evaluates the application of larger signal amplitudes despite the LED-nonlinearities to enable data transmission over long distances, evaluating the conventional and constant envelope OFDMperformances. Furthermore, it proposes a ModifiedGeneticAlgorithm(MGA) optimization procedure combined with time series Machine Learning (ML) classifiers to min imize handovers in both a digital twin-based simulation system and experimental VLC setups. The proposed handover scheme considers receiver trajectory information to reduce handover frequency while maintaining system performance within the forward error correction limit. Results demonstrate that a 9.51 Mb/s CE-OFDM system with 16-QAM subcarrier map ping in a 5MHz bandwidth outperformed a conventional OFDM system in terms of efficiency. The application of the CE-OFDM scheme in a 6m VLC link reduced the EVM from 17.5%to 10%, an improvement of approximately 43%. Additionally, the CE-OFDM-based VLC system demonstrated satisfactory performance in an 8 m link when using 4-QAM subcarrier mapping. The proposed handover scheme outperforms a power-based approach, achieving han dover reductions of 42.47% in a MISO simulation environment and up to 48.61% in a MIMO environment. In experimental scenarios with three and four transmitters, the scheme achieved reductions of 46.43% and 45.45%, respectively. These results confirm that the integration of MGAwithMLmodelseffectively minimizes handovers and improves overall VLC system per formance
Descrição
Palavras-chave
Artificial intelligence , Meta-heuristics , Handover optimization , Visible light communication