This paper presents the design and implementation of a Continuous Integration and Continuous Deployment (CI/CD) pipeline aimed at automating application containerization and deployment within Kubernetes clusters, using free or open-source tools and technologies. The CI part leverages Jenkins, integrating tools such as Git, SonarQube, and Docker while the CD part, managed by ArgoCD, pulls code from a Git repository and deploys the application into Kubernetes. Continuous monitoring is provided by Prometheus and Grafana, while Kubecost facilitates cost analysis to optimize resource utilization. Comprehensive load tests evaluate performance and stability under various workloads, offering a robust and scalable framework that enhances development efficiency, system reliability, and operational agility.

Radio map estimation (RME) aims at providing a radio frequency metric, such as the received power strength, at every location of a geographical region of interest by relying on measurements acquired at multiple positions. Although a large number of estimators have been proposed so far, their performance has been analyzed mostly on simulated data. The theoretical aspects of the RME problem as well as performance bounds remain an open problem. This paper takes a step towards filling this gap by means of a theoretical analysis of the RME problem in a free-space propagation environment. First, the complexity of the estimation problem is quantified by means of upper bounds on the spatial variability of radio maps. Second, error bounds are derived for zeroth-order and first-order interpolation estimators. The proximity coefficient, which depends proportionally on the transmitted power and inversely proportionally on the cube of the distance from the transmitters to the mapped region, is proposed to quantify the complexity of the RME problem. One of the main findings is that the error of the considered estimators is roughly proportional to this proximity coefficient. Simple numerical experiments verify the tightness of the obtained bounds.

The deployment of aerial base stations (ABSs) on unmanned aerial vehicles (UAVs) presents a promising solution for extending cellular connectivity to areas where terrestrial infrastructure is overloaded, damaged, or absent. A pivotal challenge in this domain is to decide the locations of a set of ABSs to effectively serve ground-based users. Most existing approaches oversimplify this problem by assuming that the channel gain between two points is a function of solely distance and, sometimes, also the elevation angle. In turn, this paper leverages propagation radio maps to account for arbitrary air-to-ground channel gains. This methodology enables the identification of an approximately minimal set of locations where ABSs need to be deployed to ensure that all ground terminals achieve a target service rate, while adhering to backhaul capacity limitations and avoiding designated no-fly zones. Relying on a convex relaxation technique and the alternating direction method of multipliers (ADMM), this paper puts forth a scalable solver whose computational complexity scales linearly with the number of ground terminals. Convergence is established analytically and an extensive set of simulations corroborate the merits of the proposed scheme relative to conventional methods.

Radio maps provide radio frequency metrics, such as the received signal strength, at every location of a geographic area. These maps, which are estimated using a set of measurements collected at multiple positions, find a wide range of applications in wireless communications, including the prediction of coverage holes, network planning, resource allocation, and path planning for mobile robots. Although a vast number of estimators have been proposed, the theoretical understanding of the radio map estimation (RME) problem has not been addressed. The present work aims at filling this gap along two directions. First, the complexity of the set of radio map functions is quantified by means of lower and upper bounds on their spatial variability, which offers valuable insight into the required spatial distribution of measurements and the estimators that can be used. Second, the reconstruction error for power maps in free space is upper bounded for three conventional spatial interpolators. The proximity coefficient, which is a decreasing function of the distance from the transmitters to the mapped region, is proposed to quantify the complexity of the RME problem. Numerical experiments assess the tightness of the obtained bounds and the validity of the main takeaways in complex environments.

Navigation assistance is an active research area, where one aim is to foster independent living for people with vision impairments. Despite the fact that many navigation assistants use advanced technologies and methods, we found that they did not explicitly address two essential requirements in a navigation assistant - portability and convenience. It is equally imperative in designing a navigation assistant for the visually impaired that the device is portable and convenient to use without much training. Some navigation assistants do not provide users with detailed information about the obstacle types that can be detected, which is essential to make informed decisions when navigating in real-time. To address these gaps, we propose DeepNAVI, a smartphone-based navigation assistant that leverages deep learning competence. Besides providing information about the type of obstacles present, our system can also provide information about their position, distance from the user, motion status, and scene information. All this information is offered to users through audio mode without compromising portability and convenience. With a small model size and rapid inference time, our navigation assistant can be deployed on a portable device such as a smartphone and work seamlessly in a real-time environment. We conducted a pilot test with a user to assess the usefulness and practicality of the system. Our testing results indicate that our system has the potential to be a practical and useful navigation assistant for the visually impaired.

Receiving navigation directions and information through pertinent channels is essential for people with visual impairments when they use navigation assistant systems. Different navigation systems have embraced diverse methods and modalities to give navigation instructions to users. Furthermore, various studies have been reported analyzing the modality preferences of people with visual impairments. They have reported results mainly via surveys or interviews. In this work, we have extensively studied the navigation information people with visual impairments would like to receive and how they would prefer it. Our research adopted a mixed-method design consisting of an experiment followed by a questionnaire and an interview with participants with visual impairments and with 20/20 vision but blindfolded. Unimodal (audio alone) and multimodal (audio and vibration) instruction modes and two types of audio instructions: short and descriptive have been tested. The navigation performance of participants is measured in terms of time taken and errors made while they navigate with different types and modes of instruction. The study’s findings imply no significant effect of instruction mode on time taken or errors made to complete the trials. The study could not find evidence that participants prefer descriptive instructions over short instructions. Furthermore, no evidence supports that participants prefer to hear details about the environment and obstacles during navigation. Different participants prefer different modes of instruction, which infer the importance of customized functionalities in a navigation assistant system. We also validate from our study that participants with visual impairments spent less time and made fewer errors while navigating than blindfolded participants with 20/20 vision. We believe that the results from the study will expand the knowledge of user preferences in the type and modality of instructions in the navigation assistive technology domain.

Radio maps find numerous applications in wireless communications and mobile robotics tasks, including resource allocation, interference coordination, and mission planning. Although numerous existing techniques construct radio maps from spatially distributed measurements, the locations of such measurements are predetermined beforehand. In contrast, this paper proposes spectrum surveying, where a mobile robot such as an unmanned aerial vehicle (UAV) collects measurements at a set of locations that are actively selected to obtain high-quality map estimates in a short surveying time. This is performed in two steps. First, two novel algorithms, a model-based online Bayesian estimator and a data-driven deep learning algorithm, are devised for updating a map estimate and an uncertainty metric that indicates the informativeness of measurements at each possible location. These algorithms offer complementary benefits and feature constant complexity per measurement. Second, the uncertainty metric is used to plan the trajectory of the UAV to gather measurements at the most informative locations. To overcome the combinatorial complexity of this problem, a dynamic programming approach is proposed to obtain lists of waypoints through areas of large uncertainty in linear time. Numerical experiments conducted on a realistic dataset confirm that the proposed scheme constructs accurate radio maps quickly.

Deep learning models have recently gained popularity in the research community due to their high classification success rates. In this paper, we proposed an EfficientNet-Lite based scene recognition model for scene recognition as a part of the smartphone-based navigation support application for the blind and visually impaired. We created a custom dataset with both indoor and outdoor scenes for training and testing of the model. The main objective of this work is to support people with visual impairments navigate by providing information about the scene via a smartphone application. The results from the experiment show encouraging performance from the proposed model. As a proof of concept, a prototype app was developed on the Android platform. However, the model can be implemented and deployed in any modern smartphone with good processing power.

Database is an important component in any software application, which enables efficient data management. High availability of databases is critical for an uninterruptible service offered by the application. Virtualization has been a dominant technology behind providing highly available solutions in the cloud including database, where database servers are provisioned and dynamically scaled based on demands. However, containerization technology has gained popularity in recent years because of light-weight and portability, and the technology has seen increased number of enterprises embracing containers as an alternative to heavier and resource-consuming virtual machines for deploying applications and services. A relatively new cluster-based synchronous multi-master database solution has gained popularity recently and has seen increased adoption against the traditional master-slave replication for better data consistency and high availability. This article evaluates the performance of a cluster-based high availability database deployed in containers and compares it to the one deployed in virtual machines. A popular cloud software platform, OpenStack, is used for virtual machines. Docker is used for containers as it is the most popular container technology at the moment. Results show better performance by HA Galera cluster database setup using Docker containers in most of the Sysbench benchmark tests compared to a similar setup using OpenStack virtual machines.

With the exponential growth in cloud computing, the steadily increasing amount of power consumption due to the use of physical and virtual machines is becoming a serious challenge. In this context, we report a study on optimizing the power and energy efficiency of physical and virtual machines in a cloud computing environment. The energy profile of different workloads is thoroughly investigated under different configurations. This paper presents the findings from our study which provides a good understanding of how different workloads affect power and energy efficiency of both physical and virtual machines.

With the exponential growth in cloud computing, the steadily increasing amount of power consumption due to the use of physical and virtual machines is becoming a serious challenge. In this context, we report a study on optimizing the power and energy efficiency of physical and virtual machines in a cloud computing environment. The energy profile of different workloads is thoroughly investigated under different configurations. This paper presents the findings from our study which provides a good understanding of how different workloads affect power and energy efficiency of both physical and virtual machines.