Visual understanding is a research area that bridges the gap between computer vision and natural language processing. Image captioning is a visual understanding task in which natural language descriptions of images are automatically generated using vision-language models. The transformer architecture was initially developed in the context of natural language processing and quickly found application in the domain of computer vision. Its recent application to the task of image captioning has resulted in markedly improved performance. In this paper, we briefly look at the transformer architecture and its genesis in attention mechanisms. We more extensively review a number of transformer-based image captioning models, including those employing vision-language pre-training, which has resulted in several state-of-the-art models. We give a brief presentation of the commonly used datasets for image captioning and also carry out an analysis and comparison of the transformer-based captioning models. We conclude by giving some insights into challenges as well as future directions for research in this area.
Recently, the mobile wireless communication has seen explosive growth in data traffic which might not be supported by the current Fourth Generation (4G) networks. The Fifth Generation (5G) networks will overcome this challenge by exploiting a higher spectrum available in millimeter-wave (mmwave) band to improve network throughput. The integration of the millimeter-wave communication with device-to-device communication can be an enabling 5G scheme in providing bandwidth-intensive proximity-based services such as video sharing, live streaming of data, and socially aware networking. Furthermore, the current cellular network traffic can also be offloaded by the D2D user devices thereby reducing loading at Base Stations (BSs), which would then increase the system capacity. However, the mmwave D2D communication is associated with numerous challenges, which include signal blockages, user mobility, high-computational complexity resource allocation algorithms, and increase in interuser interference for dense D2D user scenario. The paper presents review of existing channel and power allocation approaches and mathematical resource optimization solution techniques. In addition, the paper discusses the challenges hindering the realization of an effective allocation scheme in mmwave D2D communication and gives open research issues for further study.
9 January 2026 in Journal of Engineering and Applied Science
User-centric (UC) Cell-Free massive Multiple-Input Multiple-Output (CF-mMIMO) millimeter-wave (mmWave) networks are a promising solution to meet the performance requirements of next-generation wireless systems. However, maximizing energy efficiency in dense deployments remains challenging due to coordination overhead and highly dynamic propagation conditions. This work addresses uplink power control in UC CF-mMIMO networks and proposes a Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) approach trained under a centralized training and decentralized execution (CTDE) paradigm. The simulations are performed in a PyTorch library and rely on 3GPP TR 38.901 specification for the mmWave channel model over a UC architecture with 35 user equipments (UEs) and 100 distributed access points (APs). Simulation results indicate clear gains over both DRL baselines and conventional optimization methods. In particular, the proposed scheme reaches an energy efficiency of up to 380 Mbit/joule and maintains spectral efficiencies above 18 bps/Hz. Moreover, the method also preserves user-level reliability with a median minimum per-user spectral efficiency remains above 9 bps/Hz, and the Jain fairness index reaches 0.96, preventing resource starvation while maintaining strict QoS guarantees. These findings demonstrate that multi-agent cooperation enables robust and energy-efficient power control policies, paving the way for cost-effective and scalable UC CF-mMIMO deployments.
Intelligent Transport System (ITS) has emerged as the most probable technology for improved transport experience more so in environments with high vehicular density. Effective vehicular communication is however hindered by spectrum scarcity due to the already crowded licensed spectrum. This has led to the emergence of Cognitive Radio (CR) systems as a solution to the spectrum scarcity problem. A crucial component in CR is spectrum sensing. Various spectrum sensing techniques including Cyclostationary, Matched Filter, and Energy detection have been proposed and applied with varied outcomes. In this paper, the above mentioned detection techniques are discussed and an improved energy detection based cooperative spectrum sensing scheme is proposed for improved communication in vehicular ad hoc networks (VANET). The proposed scheme showed an improvement in the performance of a network which in turn could lead to more efficient utilization of spectrum.
16 August 2020 in Engineering, Technology & Applied Science Research
The current trend has seen the data capacity and traffic density increase due to the increased demand for multimedia services. Since this cannot be handled successfully by the current 4G networks, there is a need to integrate the mmWave and the Device-to-Device (D2D) communication 5G technologies to meet this increased demand and traffic density. However, there is the challenge of increased interference between dense D2D users and cellular users if D2D users are allowed to reuse the resources allocated to cellular users. This degrades the performance of the D2D users in terms of achievable data rate and Energy Efficiency (EE). The paper formulates a match theoretic resource allocation scheme to maximize the achievable D2D sum rate. In addition, an EE optimization problem is formulated for D2D users by considering the rate and power constraints. The EE optimization problem is solved by the Lagrangian dual decomposition method. The algorithms were simulated in MATLAB and the results were compared to Hungarian and heuristic optimization algorithms. The results showed that the match theoretic resource allocation is on average 1.82 times better than the Hungarian algorithm. At the same time, the match theoretic resource allocation algorithm increases fairness in resource allocation as it maintains a higher sum rate for low and high-density number of users. The proposed EE optimization algorithm improved the D2D performance by 8.2% compared to the heuristic algorithm.
With rapid advances in the emerging and pervasive field of the Internet of Things, electrical and electronic engineering students need a robust grasp of embedded computational device principles to make a proper, appreciable and relevant contribution. This necessity is more pertinent in developing countries that are technologically lagging behind their developed counterparts. This paper analyses the teaching modalities and models used to teach embedded device theory. A new triumvirate (or three person group) hybrid blended learning model is proposed and used to instruct mid-level students in the Electrical and Information Engineering Department of the University of Nairobi; on two embedded device types: a custom board using an AVR Atmega88A series micro-controller and the Raspberry Pi single-board computer. Student circuit design ability and programming confidence prior and subsequent to the proposed method were used to assess the efficacy of the triumvirate model. From the preliminary iteration of the learning model program, it was found that the model is a viable method of ensuring persistent knowledge transfer of embedded design concepts and can be developed further.