Siu-Wai Ho

Existing models for Radio Frequency (RF) and Free Space Optical (FSO) attenuations, such as the recommendations published by the International Telecommunication Union (ITU), require physical parameters along the communication channels. In practice, the weather parameters of the entire path are usually unavailable. This paper presents RF and FSO attenuation models built using machine learning algorithms and applied to empirical data. The empirical data consists of weather parameters collected at one end of the channel. Seven pairs of RF/FSO models are trained for specific weather conditions. The importance of each weather parameter is compared. RF attenuation is found to be sensitive to humidity, while FSO attenuation is closely related to scintillation. This paper shows how to obtain a pair of generic random forests that are applicable to seven specific weather conditions. The generic random forests predict the RF and FSO attenuations which have a joint distribution similar to the empirically observed distributions. They preserve the correlation between the RF and FSO attenuations as measured by the correlation coefficient and mutual information. When applied to empirical data, the generic random forests outperform the ITU models and models constructed by linear regression with interaction, both in terms of Root-Mean-Square-Error and R-squared.

There is a growing demand for indoor positioning systems (IPSs) in a wide range of applications. However, traditional solutions such as GPS face many technical challenges. In recent years, a promising alternative has been emerging, the visible light communication (VLC)-based IPS, which offers a combination of high accuracy, low cost, and energy efficiency. This article presents a comprehensive review of VLC-based IPSs, providing a tutorial-like overview of the system. It begins by comparing various positioning systems and providing background information on their inherent limitations. Experimental results have demonstrated that VLC-based systems can achieve localization accuracy to within 10 cm in controlled environments. The mechanisms of VLC-based IPSs are then discussed, including a comprehensive examination of their performance metrics and underlying assumptions. The complexity, operating range, and efficiency of VLC-based IPSs are examined by analyzing factors such as channel modeling, signal processing, and localization algorithms. To optimize VLC-based IPSs, various strategies are explored, including the design of efficient modulation schemes, the development of advanced encoding and decoding algorithms, the implementation of adaptive power control, and the application of state-of-the-art localization algorithms. In addition, system parameters are carefully examined. These include LED placement, receiver sensitivity, and transmit power. Their impact on energy efficiency and localization accuracy is highlighted. Altogether, this paper serves as a comprehensive guide to VLC IPSs, providing in-depth insights into their vast potential and the challenges that they present.

There is a growing demand for indoor positioning systems (IPS) in a wide range of applications. However, traditional solutions such as GPS face many technical challenges. In recent years, a promising alternative has been emerging: IPS based on visible light communication (VLC), which offers a combination of which is highly accurate, low cost and energy efficient. This article presents a comprehensive review of VLC-based IPS, providing a tutorial-like overview of the system. It begins by comparing various positioning systems and providing background information on their inherent limitations. The mechanisms of VLC-based IPS are then discussed, including a comprehensive examination of their performance metrics and underlying assumptions. The complexity and efficiency of VLC-based IPS is examined by analyzing factors such as channel modeling, signal processing, and localization algorithms. To optimize the system, a variety of strategies are explored. These include the design of efficient modulation schemes, the development of advanced encoding and decoding algorithms, the implementation of adaptive power control, and the use of state-of-the-art localization algorithms. Additionally, system parameters such as LED placement, receiver sensitivity, and transmit power are carefully examined to highlight their impact on energy efficiency and localization accuracy. Altogether, this paper serves as a comprehensive guide to VLC IPS, providing in-depth insights into its vast potential and the challenges it presents.

This paper proposes a novel indoor positioning system that can simultaneously track multiple battery-free tags with centimeter accuracy. Our system uses a camera located near a light source to track tags with special designs. Our tags have unique fiducial markers printed on prismatic retroreflective material instead of paper used in traditional systems. Our system has achieved average position errors of less than 2.9 cm and 7.1 cm for perpendicular distances between the camera and tags of 3.28 m to 7.48 m, respectively. When the tags move at a speed of 1.12 m/s, our system remains unaffected in terms of positioning accuracy. Our tags are more robust to changes in lighting conditions than traditional paper tags because our tags can be located in five different scenarios, including a dark environment where traditional paper tags cannot be detected. Experimental results show that the working range of our tag is approximately twice that of paper tags, and the position error is not significantly affected by the ambient lighting conditions. Our system can be implemented using smartphones without any expensive add-on equipment.

A hybrid Radio Frequency/Free Space Optical (RF/FSO) communication system exploits the benefits of channel physical layer diversity to improve data rate and link availability. The realisable benefits depend on the respective channel attenuations and their level of independence in the presence of degrading effects. The impact of weather on link attenuations is reported in this paper based on empirical data obtained from a hybrid RF/FSO system where the physical paths were near-coincident and the RF was in the E band. The system was operated in six cities (across five countries) around the world and approximately 15,000 hours of logged data, including weather, was available. This paper is possibly the first to use empirical data to investigate the correlation between the RF and FSO channels across a wide range of weather conditions including dust storms, fog, drizzle, rain, snow, showers and clear skies. It reports 1) the distributions of RF attenuation under different FSO channel states; 2) the joint distribution of the RF and FSO attenuations; and 3) the Pearson correlation coefficient and mutual information. Our empirical data illustrate a negative non-linear correlation during dust storm and a positive linear correlation during fog and snow.

Existing models for Radio Frequency (RF) and Free Space Optical (FSO) attenuations, such as the recommendations published by the International Telecommunication Union (ITU), require physical parameters along the communication channels. In practice, the weather parameters of the entire path are usually unavailable. This paper presents RF and FSO attenuation models built using machine learning algorithms and applied to empirical data. The empirical data consists of weather parameters collected at one end of the channel. Seven pairs of RF/FSO models are trained for specific weather conditions. The importance of each weather parameter is compared. RF attenuation is found to be sensitive to humidity, while FSO attenuation is closely related to scintillation. This paper shows how to obtain a pair of generic random forests that are applicable to seven specific weather conditions. The generic random forests predict the RF and FSO attenuations which have a joint distribution similar to the empirically observed distributions. They preserve the correlation between the RF and FSO attenuations as measured by the correlation coefficient and mutual information. When applied to empirical data, the generic random forests outperform the ITU models and models constructed by linear regression with interaction, both in terms of Root-Mean-Square-Error and R-squared.

A hybrid Radio Frequency/Free Space Optical (RF/FSO) communication system exploits the benefits of channel physical layer diversity to improve data rate and link availability. The realisable benefits depend on the respective channel attenuations and their correlation in the presence of degrading effects. The impact of weather on link attenuations is reported in this paper based on empirical data obtained from a hybrid RF/FSO system where the physical paths were near-coincident and the RF was in the E band. The system was operated in six cities (across five countries) around the world and approximately 15,000 hours of logged data, including weather, was available. The weather conditions encountered included dust storms, fog, drizzle, rain, snow, showers and clear skies. This paper is possibly the first to use empirical data to investigate the correlation between the RF and FSO channels across this range of weather conditions. It reports 1) the distributions of RF attenuation under different FSO channel states; 2) the joint distribution of the RF and FSO attenuations; and 3) the Pearson correlation coefficient and mutual information. A negative non-linear correlation was observed during dust storm and a positive linear correlation was observed during fog and snow.

This paper provides a comprehensive review of the literature concerning the utilization of Natural Language Processing (NLP) techniques, with a particular focus on transformer-based large language models (LLMs) trained using Big Code, within the domain of AI-assisted programming tasks. LLMs, augmented with software naturalness, have played a crucial role in facilitating AI-assisted programming applications, including code generation, code completion, code translation, code refinement, code summarization, defect detection, and clone detection. Notable examples of such applications include the GitHub Copilot powered by OpenAI's Codex and DeepMind AlphaCode. This paper presents an overview of the major LLMs and their applications in downstream tasks related to AI-assisted programming. Furthermore, it explores the challenges and opportunities associated with incorporating NLP techniques with software naturalness in these applications, with a discussion on extending AI-assisted programming capabilities to Apple's Xcode for mobile software development. This paper also presents the challenges of and opportunities for incorporating NLP techniques with software naturalness, empowering developers with advanced coding assistance and streamlining the software development process.

This paper provides a comprehensive review of the literature concerning the utilization of Natural Language Processing (NLP) techniques, with a particular focus on transformer-based large language models (LLMs) trained using Big Code, within the domain of AI-assisted programming tasks. LLMs, augmented with software naturalness, have played a crucial role in facilitating AI-assisted programming applications, including code generation, code completion, code translation, code refinement, code summarization, defect detection, and clone detection. Notable examples of such applications include the GitHub Copilot powered by OpenAI’s Codex and DeepMind AlphaCode. This paper presents an overview of the major LLMs and their applications in downstream tasks related to AI-assisted programming. Furthermore, it explores the challenges and opportunities associated with incorporating NLP techniques with software naturalness in these applications, with a discussion on extending AI-assisted programming capabilities to Apple’s Xcode for mobile software development. This paper also presents the challenges of and opportunities for incorporating NLP techniques with software naturalness, empowering developers with advanced coding assistance and streamlining the software development process.

In some visible light communication (VLC) systems, transmitters are light sources which support not only illumination but also information broadcast and positioning. Since transmit signals in these systems inevitably interfere with each other at the receiver side, extra devices may be added to synchronize the transmitters, but this impairs an appealing advantage in VLC systems that existing lighting infrastructure can be simply reused. This paper proposes a novel multiple access scheme to realize decentralized asynchronous VLC systems such that both the transmitters and receivers are asynchronous. For a system with $N$ transmitters, two unique codewords of $L$ chips are allocated to each transmitter where $L$ is a prime number larger than $4N$ . Those codewords are designed to help maintain a constant transmitted average power and enhance channel estimation, as well as conveying data. Our scheme enables the receiver to obtain one channel estimate and one data symbol per $L$ chips. Simulation results show that the proposed scheme significantly outperforms other schemes in the literature in terms of bit error rate and system throughput.