Memory Resources Research Articles

An examination of the effects of eye-tracking on behavior in psychology experiments.

Eye-tracking is emerging as a tool for researchers to better understand cognition and behavior. However, it is possible that experiment participants adjust their behavior when they know their eyes are being tracked. This potential change would be considered a type of Hawthorne effect, in which participants alter their behavior in response to being watched and could potentially compromise the outcomes and conclusions of experimental studies that use eye tracking. We examined whether eye-tracking produced Hawthorne effects in six commonly used psychological scales and five behavioral tasks. The dependent measures were selected because they are widely used and cited and because they involved measures of sensitive topics, including gambling behavior, racial bias, undesirable personality characteristics, or because they require working memory or executive attention resources, which might be affected by Hawthorne effects. The only task where Hawthorne effects manifested was the mixed gambles task, in which participants accepted or rejected gambles involving a 50/50 chance of gaining or losing different monetary amounts. Participants in the eye-tracking condition accepted fewer gambles that were low in expected value, and they also took longer to respond for these low-value gambles. These results suggest that eye-tracking is not likely to produce Hawthorne effects in most common psychology laboratory tasks, except for those involving risky decisions where the probability of the outcomes from each choice are known.

Approximate and Memorize (A&M) : Settling opposing views in replay-based continuous unsupervised domain adaptation

Continuous Unsupervised Domain Adaptation (CUDA) can alleviate deep learning models’ performance degradation on out-of-distribution data. However, low stability, the erosion of past knowledge while adapting to new domains, remains a major challenge. Traditional proposed solutions such as (1) Approximation, or (2) Memorization fail to harmonize system’s computational load, memory resources and stability. In response, we introduce Approximate and Memorize (A&M). Compared to traditional approximation methods, A&M utilizes factorized generative models to mitigate modal collapse, offering improved computational efficiency and training stability. Moreover, compared to traditional memorization approaches, A&M boosts memory efficiency by learning compressed representations of past information rather than storing it in the raw format. A&M shows 50% less forgetting for problems with constrained memory and an extended number of domains, making A&M ideal for real-world systems due to its scalable and parallel design.

A comparative study of large-scale and lightweight convolutional neural networks for ImageNet classification

In the field of convolution neural networks (CNNs), many impressive architectures have been published in recent years. These can be roughly divided into two groups: large-scale models and lightweight models. These large models are characterized by many trainable weights and complex network structures, offering them strong effectiveness in various computer vision tasks, and have become essential components of many modern visual recognition systems. These lightweight CNNs are designed to maintain high performance with limited memory and computational resources. They are highly efficient in terms of inference time and resource utilization, so that particularly suitable for mobile and edge computing devices. This work focuses on some prominent models based on the ImageNet database and explores the reasons for their frameworks success. By analyzing these models, a trend could be identified in the development of CNN models: Reasonably increasing the scale of the model and utilizing suitable frameworks can both improve accuracy and efficiency.

Real-Time Human Activity Recognition on Embedded Equipment: A Comparative Study

As living standards improve, the growing demand for energy, comfort, and health monitoring drives the increased importance of innovative solutions. Real-time recognition of human activities (HAR) in smart homes is of significant relevance, offering varied applications to improve the quality of life of fragile individuals. These applications include facilitating autonomy at home for vulnerable people, early detection of deviations or disruptions in lifestyle habits, and immediate alerting in the event of critical situations. The first objective of this work is to develop a real-time HAR algorithm in embedded equipment. The proposed approach incorporates the event dynamic windowing based on space-temporal correlation and the knowledge of activity trigger sensors to recognize activities in the case of a record of new events. The second objective is to approach the HAR task from the perspective of edge computing. In concrete terms, this involves implementing a HAR algorithm in a “home box”, a low-power, low-cost computer, while guaranteeing performance in terms of accuracy and processing time. To achieve this goal, a HAR algorithm was first developed to perform these recognition tasks in real-time. Then, the proposed algorithm is ported on three hardware architectures to be compared: (i) a NUCLEO-H753ZI microcontroller from ST-Microelectronics using two programming languages, C language and MicroPython; (ii) an ESP32 microcontroller, often used for smart-home devices; and (iii) a Raspberry-PI, optimizing it to maintain accuracy of classification of activities with a requirement of processing time, memory resources, and energy consumption. The experimental results show that the proposed algorithm can be effectively implemented on a constrained resource hardware architecture. This could allow the design of an embedded system for real-time human activity recognition.

CAFE: Towards Compact, Adaptive, and Fast Embedding for Large-scale Recommendation Models

Recently, the growing memory demands of embedding tables in Deep Learning Recommendation Models (DLRMs) pose great challenges for model training and deployment. Existing embedding compression solutions cannot simultaneously meet three key design requirements: memory efficiency, low latency, and adaptability to dynamic data distribution. This paper presents CAFE, a Compact, Adaptive, and Fast Embedding compression framework that addresses the above requirements. The design philosophy of CAFE is to dynamically allocate more memory resources to important features (called hot features), and allocate less memory to unimportant ones. In CAFE, we propose a fast and lightweight sketch data structure, named HotSketch, to capture feature importance and report hot features in real time. For each reported hot feature, we assign it a unique embedding. For the non-hot features, we allow multiple features to share one embedding by using hash embedding technique. Guided by our design philosophy, we further propose a multi-level hash embedding framework to optimize the embedding tables of non-hot features. We theoretically analyze the accuracy of HotSketch, and analyze the model convergence against deviation. Extensive experiments show that CAFE significantly outperforms existing embedding compression methods, yielding 3.92% and 3.68% superior testing AUC on Criteo Kaggle dataset and CriteoTB dataset at a compression ratio of 10000x. The source codes of CAFE are available at GitHub.

The Physical and Psychosocial Impact of Fatigue among Patients with Sjogren's Syndrome: A Systematic Review.

Background: Primary Sjögren's syndrome (pSS) is a complex autoimmune disorder characterized by organ-specific symptoms in the salivary and lacrimal glands, as well as systemic manifestations. Fatigue, a prominent aspect, significantly influences the overall quality of life for individuals with pSS. Methods: This review seeks to evaluate the impact of fatigue by exploring its consequences, potential causes, and effects on physical and psychological well-being, while also investigating its management strategies. Following the "Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)" guidelines, our systematic literature review involved a five-step algorithm. Initially identifying 78 articles in reputable international medical databases, we applied eligibility criteria and removed duplicates, resulting in 19 articles for qualitative synthesis. Results: This review delves into the predictive factors for heightened fatigue in pSS, encompassing rheumatoid factor levels, erythrocyte sedimentation rate, and immunoglobulin G levels. Sleep disturbances, specifically nighttime pain and nocturia, emerged as determinants of persistent daytime fatigue. Cognitive impairment in pSS involves deteriorations in global memory, executive functioning, and attentional resources. Furthermore, functional limitations in pSS impact patients' quality of life. Conclusions: The significance of fatigue in pSS, its consequences, and profound influence on the quality of life necessitate further research for a more comprehensive understanding of this complex issue.

Enhancing Communication Security an In-Vehicle Wireless Sensor Network

Confronting the challenges of securing communication in-vehicle wireless sensor networks demands innovative solutions, particularly as vehicles become more interconnected. This paper proposes a tailored communication security framework for in-vehicle wireless sensor networks, addressing both scientific and technical challenges through effective encryption methods. It segments the local vehicle network into independent subsystems communicating via encrypted and authenticated tunnels, enhancing automotive system safety and integrity. The authors introduce a process for periodic cryptographic key exchanges, ensuring secure communication and confidentiality in key generation without disclosing parameters. Additionally, an authentication technique utilizing the sender’s message authentication code secures communication tunnels, significantly advancing automotive cybersecurity and interconnectivity protection. Through a series of steps, including key generation, sending, and cryptographic key exchange, energy costs were investigated and compared with DTLS and TLS methods. For cryptographic security, testing against brute-force attacks and analysis of potential vulnerabilities in the AES-CBC 128 encryption algorithm, HMAC authentication, and HKDF key derivation function were carried out. Additionally, an evaluation of the memory resource consumption of the DTLS and TLS protocols was compared with the proposed solution. This work is crucial for mitigating risks associated with in-vehicle communication compromises within smart cities.

Efficient Energy Level Calculations in InP 2D-Quantum Box with Two Distinct Potentials Using the Sparse Numerov Method

In this study, energy level calculations for an InP 2D quantum box structure with two distinct (infinite potential power-exponential) potential potentials have been conducted using the sparse Numerov method. The 2D Schrödinger equation has been transformed in accordance with the sparse Numerov approach, followed by the creation of the solution matrix employing appropriate finite difference expressions. A comparative analysis of calculation results has been performed with respect to CPU time, memory usage, and ground state energy for both O(h^4) and O(h^6) accuracy. The suitability of the sparse Numerov method for 2D nanostructures has been thoroughly discussed. The results revealed that the sparse Numerov approach yields physically meaningful and rational outcomes in the InP 2D quantum box structure. Importantly, it demands significantly lower CPU time and memory resources compared to the classical Numerov method, emphasizing its practical applicability in this context.

Efficient LiDAR-Trajectory Affinity Model for Autonomous Vehicle Orchestration

Computation and memory resource management strategies are the backbone of continuous object tracking in intelligent vehicle orchestration. Multi-object tracking generates enormous measurements of targets and extended object positions using light detection and ranging (Lidar) sensors. Designing an adequate object-tracking system is a global challenge because of dynamic object detection and data association uncertainties during scene understanding. In this regard, we develop an intelligent multi-objective tracking (IMOT) system with a novel measurement model, called the box data association inflate (BDAI) model, to assess each target’s object state and trajectory without noise by using the Bayesian approach. The box object filter method filters ambiguous detection responses during data association. The theoretical proof of the box object filter is derived based on binomial expansion. Prognosticating a lower-dimension object than the original point object reduces the computational complexity of vehicle orchestration. Two datasets (NuScenes dataset and our lab dataset) are considered during the simulations, and our approach measures the kinematic states adequately with reduced computation complexity compared to state-of-the-art methods. The simulation outcomes show that our proposed method is effective and works well to detect and track objects. The NuScenes dataset contains 28130 samples for training, 6019 examples for validation and 6008 samples for testing. IMOT achieves 58.09% tracking accuracy and 71% mAP with 5 ms pre-processing time. The Jetson Xavier NX consumes 49.63% GPU and 9.37% average power and exhibits 25.32 ms latency compared to other approaches. Our system trains a single pair frame in 169.71 ms with affinity estimation time of 12.19 ms, track association time of 0.19 ms and mATE of 0.245 compared to state-of-the-art approaches

FPHO: Fractional Pelican Hawks optimization based container consolidation in CaaS cloud

AbstractContainers in cloud computing provide a logical packaging technique for applications to be isolated from the computing environment in which they actually execute, allowing for efficient sharing of memory, processor, storage, and network resources at the Operating System (OS) level. Since they are so compact, container‐based clouds have recently gained significant popularity. In order to maximize resource usage and minimize energy consumption, the container consolidation technique is widely employed in the cloud environment. This work introduces container consolidation in cloud computing that exploits Fractional Pelican Hawks Optimization (FPHO). In Container as a Service (CaaS) model, containers are placed in the Virtual Machines (VMs), and virtual machines are hosted in Physical Machines (PMs) or servers. The proposed method for container consolidation consists of two modules, namely, the host status module and the consolidation module. In the host status module, the PM's load is predicted using Long Short Term Memory (LSTM) and checked whether the PM is overloaded or underloaded using a threshold. If it is overloaded, the container selection algorithm is performed, and the migration list is also generated. In the consolidation module, the created migration list which is employed for the destination list to be created by an overloaded destination selector. In the same way, the underloaded list is also generated by the underloaded destination selector. Finally, the container and VM migration is carried out by considering the multi‐objectives such as predicted load, migration cost, resource utilization, energy consumption, network, and bandwidth which are optimally selected by the proposed FHPO. Here, FHPO is the combination of Fractional Pelican Optimization (FPO) and Fire Hawk Optimizer (FHO). The designed model achieved the measures with minimum energy consumption, resource utilization, Service Level Agreement (SLA), and Makespan as 0.066, 0.019, 0.054, and 0.066, respectively for setup one.

잠재프로파일 분석을 통한 대학 신입생의 대학생활적응 잠재집단분류와 자기조절학습 전략 영향요인 탐색

Objectives This study aimed to identify latent group types of college life adaptation among university freshmen and to analyze the factors of self-regulated learning strategies that influence each group type.
 Methods In accordance with the research objectives, latent groups for types of college life adaptation were classified among 1,265 freshmen at A University located in the metropolitan area. Factors influencing each latent group regarding self-regulated learning strategies were identified. Responses from the Self-regulated Learning Test (SLT) were analyzed using latent profile analysis and multinomial logistic regression.
 Results The analysis of freshman students' adaptation to college life revealed that they can be divided into three main groups: ‘Maladaptation’, ‘Marginal’, and ‘Adaptation’. Furthermore, among the various sub-elements of self-directed learning strategies, it was found that students with higher abilities in memory strategies, emotional regulation, and resource utilization are more likely to belong to the group that adapts well to college life. Notably, students with higher metacognitive abilities have a higher probability of belonging to the ‘Adaptation’ group, which shows the highest level of college life adaptation.
 Conclusions These findings suggest the need to consider types of college life adaptation in understanding latent group profiles and predictive factors. By addressing variables of self-regulated learning strategies that influence college life adaptation, this study provides implications for necessary interventions to assist freshmen in adapting to college life.

Gaining a deeper understanding of the deep cloze reading comprehension test: examining potential contributors and consequences

AbstractThe deep cloze test was developed by Jensen and Elbro (Read Writ Interdiscip J 35(5):1221–1237, 2022. https://doi.org/10.1007/s11145-021-10230-w) to assess reading comprehension at the level of global situational understanding. In two independent studies, we examined potential contributors to students’ scores on the deep cloze reading comprehension test, as well as the predictability of students’ scores on this measure for their course achievement and integrated text understanding measured with an open-ended written comprehension assessment. Results showed that students’ language background, word recognition skills, and working memory resources explained unique portions of the variance in students’ scores on the deep cloze reading comprehension test. Further, scores on this test were positively correlated with students’ course achievement and uniquely predicted their integrated text understanding when language background, working memory, and prior topic knowledge were controlled for. Taken together, our findings support an interpretation of the deep cloze reading comprehension test as an effective and efficient measure of situation level understanding that draws on language skills, word level processes, and working memory resources and also can be used to predict students’ performance on important criterial tasks requiring deeper level understanding.

Spatial Transformation Accelerator with Parallel Data Access Scheme for Sample Reconstruction

Spatial image transformation is a commonly used component in many image processing pipelines. It enables the correction of optical distortions, image registration onto a common reference plane, electronic image stabilisation, digital zoom, video mosaicking, etc. With the growing tendency to embed image processing in low-power devices, attaining an efficient transformation solution becomes increasingly decisive. Furthermore, interpolation is the key operation in achieving the high quality of the transformed data from the original data. Fortunately, different implementations have already seen several efficiency improvements in recent years. However, interpolation relies on sampling a set of neighbouring points from memory, which has yet to be addressed efficiently for smaller computational platforms with limited memory resources. In this work, we derive a generic mathematical model and circuit design principles for the spatial transformation accelerator design for N-dimensional data. Furthermore, we present an efficient simultaneous access scheme for high-quality signal reconstruction. Finally, the introduced ideas are verified in field programmable gate arrays using one-dimensional and two-dimensional data transformation use cases. The presented solution is able to transform images with sizes ranging from 256 × 256 to 8192 × 8192 and achieves a transfer rate of 275 frames per second with 512 × 512 images.

A lightweight hardware implementation of CRYSTALS-Kyber

The security of cryptographic algorithms based on integer factorization and discrete logarithm will be threatened by quantum computers in future. Since December 2016, the National Institute of Standards and Technology (NIST) has begun to solicit post-quantum cryptographic (PQC) algorithms worldwide. CRYSTALS-Kyber was selected as the standard of PQC algorithm after 3 rounds of evaluation. Meanwhile considering the large resource consumption of current implementation, this paper presents a lightweight architecture for ASICs and its implementation on FPGAs for prototyping. In this implementation, a novel compact modular multiplication unit (MMU) and compression/decompression module is proposed to save hardware resources. We put forward a specially optimized schoolbook polynomial multiplication (SPM) instead of number theoretic transform (NTT) core for polynomial multiplication, which can reduce about 74% SLICE cost. We also use signed number representation to save memory resources. In addition, we optimize the hardware implementation of the Hash module, which cuts off about 48% of FF consumption by register reuse technology. Our design can be implemented on Kintex-7 (XC7K325T-2FFG900I) FPGA for prototyping, which occupations of 4777/4993 LUTs, 2661/2765 FFs, 1395/1452 SLICEs, 2.5/2.5 BRAMs, and 0/0 DSP respective of client/server side. The maximum clock frequency can reach at 244 ​MHz. As far as we know, our design consumes the least resources compared with other existing designs, which is very friendly to resource-constrained devices.

An Empirical Evaluation on the Effect Refactoring Code Smells Mobile Applications Android with ASATs on Resource Usage

The Application is closely connected to mobile devices designed for many people and has maintenance, but even maintenance can contain violations such as code smells that effect non-functional requirements, specifically the use of CPU and memory resources. When the software has a rapid use of resources, it gives rise to the phenomenon that the user may switch or uninstall the software. The solution to this phenomenon is to explore resource-related code smells and fix them by refactoring. Developments to explore code smells came with ASATs, namely SonarQube, which is already being used by 85,000 organizations to speed up analyzing code in software. This topic is related to code smells, and the research objective is to analyze and compare the performance of the original versions and single or cumulative refactored versions of Android mobile software using the Design Research Methodology (DRM) approach. Types of code smells are represented based on the classification on SonarQube, namely Blocker, Critical, Major, and Minor, with code smells such as HashMap Usage, Member Ignoring Method, and Slow Loop. Aspects tested include Fixed Detection Ratio (FDR), improvement, CPU, and memory usage. Based on the results of the research, it shows the depreciation of code smells which is proven to significantly increase CPU performance in a single refactoring, namely Member Ignoring Method and Critical by 7.7% and 9.90%, respectively. Moreover, single refactoring offers developers advantages reducing high costs, diminished exertion, and truncated maintenance duration. However, the cumulative refactoring occasionally endeavors hold the potential be high improvements.

CrossPrune: Cooperative pruning for camera–LiDAR fused perception models of autonomous driving

Deep neural network pruning is effective in enabling high-performance perception models to be deployed on autonomous driving platforms with limited computation and memory resources. With their rapid development, state-of-the-art autonomous driving perception (ADP) models have advocated the use of multimodal sensors to extract diverse feature categories. However, existing pruning studies in the ADP area focus on single-modal models and neglect multimodal models. Compared with conventional pruning, multimodal pruning presents a new type of redundancy, namely modal-wise redundancy that is caused by multimodal branches extracting similar perception information. When a specific type of information is extracted by more than one modal branch, although this extraction is deemed essential from an individual single-modal standpoint, it becomes redundant when viewed from a modal-wise perspective. Therefore, modal branches must be handled cooperatively to eliminate modal-wise redundancy while concurrently preserving the original perception accuracy. Building on this, we propose CrossPrune, a modal cooperative pruning framework designed for camera–LiDAR fused perception in autonomous driving. The primary objective of CrossPrune is to effectively eliminate redundancy and achieve nondestructive pruning for multimodal ADP models. This was accomplished by approaching the problem as a multi-objective optimization task, encompassing both weight pruning and the restriction of feature distortions caused by pruning. Experiments conducted on the nuScenes and KITTI datasets demonstrated that CrossPrune attained superior pruning ratios while minimizing accuracy loss, surpassing the performance of the baselines. The key results indicated that the proposed CrossPrune achieved relative improvements of 9.6% in mAP and 11.5% in NDS under 89.8% pruning sparsity.

Efficient image restoration with style-guided context cluster and interaction

Recently, convolutional neural networks (CNNs) and vision transformers (ViTs) have emerged as powerful tools for image restoration (IR). Nonetheless, they encountered some limitations due to their characteristics, such as CNNs sacrificing global reception and ViTs requiring large memory and graphics resources. To address these limitations and explore an alternative approach for improved IR performance, we propose two clustering-based frameworks for general IR tasks, which are style-guided context cluster U-Net (SCoC-UNet) and style-guided clustered point interaction U-Net (SCPI-UNet). The SCoC-UNet adopts a U-shaped architecture, comprising position embedding, Encoder, Decoder, and reconstruction block. Specifically, the input low-quality image is viewed as a set of unorganized points, each of which is first given location information by the continuous relative position embedding method. These points are then fed into a symmetric Encoder and Decoder which utilize style-guided context cluster (SCoC) blocks to extract potential context features and high-frequency information. Although SCoC-UNet has obtained decent performance for image restoration, its SCoC block can only capture connectivity at points within the same cluster, which may ignore long-range dependencies in different clusters. To address this issue, we further propose a SCPI-UNet based on SCoC-UNet, which leverages a style-guided clustered point interaction (SCPI) block in place of the SCoC block. The SCPI block utilizes a cross-attention mechanism to establish the connections of feature points between different clusters. Extensive experimental results demonstrate that the proposed SCoC-UNet and SCPI-UNet can handle several typical IR tasks (i.e., JPEG compression artifact reduction, image denoising, and super-resolution) and achieve superior quantitative and qualitative performance over some state-of-the-art methods.

An efficient and accurate semi-analytical matching technique for waveguide-fed antennas

Several RF and microwave radiating devices, such as horn antennas, Fabry–Perot cavity antennas, and aperture-fed focusing devices, are excited through rectangular waveguides. The impedance matching of the overall system (from the waveguide feed to the radiating aperture) is a task of crucial importance that is often addressed by means of brute-force parameter-sweep full-wave analyses or blind optimization algorithms. In both cases, a significant amount of memory and time resources are required. For this purpose, we propose here a simple, yet effective solution, which only requires a single full-wave simulation and a semi-analytical procedure. The former is used to retrieve the antenna input impedance at the end of the waveguide port excitation. The semi-analytical procedure consists in a transmission-line equivalent circuit that models two waveguide discontinuities (namely two capacitive irises) within the waveguide section, whose position and geometric features are finely tuned to obtain a satisfactory impedance matching around the working frequency. The proposed method is shown to be effective in diverse and attractive application-oriented contexts, from the impedance matching of a Fabry–Perot cavity antenna to that of a wireless near-field link between two aperture-fed focusing devices. A remarkable agreement between full-wave simulations and numerical results is found in all cases. Thanks to its versatility, simplicity, and a rather low demand of computational resources, the proposed approach may become an essential tool for the effective design of waveguide-fed antennas.

Eciton: Very Low-power Recurrent Neural Network Accelerator for Real-time Inference at the Edge

This article presents Eciton, a very low-power recurrent neural network accelerator for time series data within low-power edge sensor nodes, achieving real-time inference with a power consumption of 17 mW under load. Eciton reduces memory and chip resource requirements via 8-bit quantization and hard sigmoid activation, allowing the accelerator as well as the recurrent neural network model parameters to fit in a low-cost, low-power Lattice iCE40 UP5K FPGA. We evaluate Eciton on multiple, established time-series classification applications including predictive maintenance of mechanical systems, sound classification, and intrusion detection for IoT nodes. Binary and multi-class classification edge models are explored, demonstrating that Eciton can adapt to a variety of deployable environments and remote use cases. Eciton demonstrates real-time processing at a very low power consumption with minimal loss of accuracy on multiple inference scenarios with differing characteristics, while achieving competitive power efficiency against the state-of-the-art of similar scale. We show that the addition of this accelerator actually reduces the power budget of the sensor node by reducing power-hungry wireless transmission. The resulting power budget of the sensor node is small enough to be powered by a power harvester, potentially allowing it to run indefinitely without a battery or periodic maintenance.

Unmanned aerial vehicle location systems. Antenna Arrays

Radars remain the only all–weather location systems, which is especially important in difficult conditions of their use. Antenna array technologies remain critical for the development of dynamic areas of technology and, first of all, UAVs location systems. The development of location systems and antenna arrays for UAVs has been going on for a long time. Radars and antenna arrays installed on medium– and large–class UAVs differ little from airplane and helicopter systems. Among UAVs, the maximum number of UAVs used are UAVs of minimal size, weight and cost. Due to their high maneuverability, minimal visibility and a large number of tasks to be solved. Therefore, radars are subject to minimum size, weight, dimensions and consumption requirements at maximum range. The widespread use of small UAVs has led to the fact that their radars and antenna arrays are on a par with automotive ones, when the possibilities for their placement are significantly limited, even more so than in cars. For the latter, the problem of miniaturization appeared earlier, but not as severely as in UAVs. Antenna arrays are subject to requirements for minimizing dimensions, weight and consumption. Nanotechnology provides this. It's safe to say that the time has come for large–scale solutions to complex problems in antenna systems. Depending on the type of a UAV, it is possible to use a number of methods for constructing antenna arrays: known and proven – linear and fractal; the most modern – metastructures; widely used – tiled; new: flexible and three–dimensional printed; a combination of several methods. Currently, the most attractive areas for the development of antenna arrays for UAVs are: scaling – growing a large number of similar structures to create large apertures; spacing antenna arrays across the surface of the carrier for viewing both forward, sideways and backward, as well as down and up. The development of silicon tiled antenna arrays has led to the realization of very high operating frequencies (90–100 GHz) on the most affordable process technologies (130–180 nm), and the antenna array is fabricated as a chip. The optimal option for constructing antenna arrays is to ensure the reception, conversion and preprocessing of input signals directly in the antenna system in real time, unloading the radar for solving problems that require powerful processing. Post–processing allows you to obtain a more accurate solution to the problem. The combination of pre–processing and post–processing, supported by advanced software and powerful artificial intelligence, expands the capabilities and improves the parameters of the location system. It is important that artificial intelligence ensures the autonomous movement of a UAV, although its use requires large consumption and memory resources. The integrated location system absorbed for the most part the capabilities of a large stationary radar. At the same time, minimum dimensions, weight and consumption are ensured to solve the necessary modern problems. Several radars with their own antenna arrays can be installed on a UAV, which operate in the internal network, complementing each other.