This letter presents a heterogeneous Unmanned Aerial Vehicle (UAV)-based robotic system for real-time barcode detection and scanning using Convolutional Neural Networks (CNN). The proposed approach improves the UAV's localization scanned barcodes as landmarks in real warehouse with low-light conditions. Instead of standard overlapping snake-based grid (OSBG) trajectory, we implement novel flight-path optimization based on locations. reduces time stocktaking decreases number mistakes scanning.

The paper focuses on the development of autonomous robot UltraBot to reduce COVID-19 transmission and other harmful bacteria viruses. motivation behind research is develop such a that capable performing disinfection tasks without use sprays chemicals can leave residues, require airing room afterward for long time, cause corrosion metal structures. technology has potential offer most optimal performance along with taking care people, keeping them from getting under UV-C radiation. highlights...

We present a novel high-precision UAV localization system for interconnection between two collaborative robots, i.e., unmanned ground robot (UGR) and aerial vehicle (UAV) capable of autonomous navigation precise in an indoor environment. Based on our we have achieved robust landing the moving using fusion 2D LIDAR sensors, camera, ultrasonic localization. In addition, is accurate high-altitude flights (up to 15 m) relative robot. Localization based developed adaptive active IR marker achieve...

This paper suggests a novel method for customer behavior analytics and demand distribution based on Radio Frequency Identification (RFID) stocktaking. Existing solutions lack applicability to real-life situations in retailing, which may result unobservable loss of sales. The proposed solution provides new parameters the retailer using mobile robot autonomous stocktaking RFID-equipped shopping rooms. Built models depict location-related dependencies, most least purchasable areas store,...

Existing Simultaneous Localization and Mapping (SLAM) approaches are limited in their scalability due to growing map size long-term robot operation. Moreover, processing such maps for localization planning tasks leads the increased computational resources required onboard. To address problem of memory consumption operation, we develop a novel real-time SLAM algorithm, MeSLAM, that is based on neural field implicit representation. It combines proposed global mapping strategy, including...

Optimal motion planning is one of the most critical problems in mobile robotics. On hand, classical sampling-based methods propose asymptotically optimal solutions to this problem. However, these planners cannot achieve smooth and short trajectories reasonable calculation time. other optimization-based are able generate plain a variety scenarios, including dense human crowd. modern use precomputed signed distance function for collision loss estimation, it limits application general...

In the proposed study, we describe an approach to improving computational efficiency and robustness of visual SLAM algorithms on mobile robots with multiple cameras limited power by implementing intermediate layer between pipeline. this layer, images are classified using a ResNet18-based neural network regarding their applicability robot localization. The is trained six-camera dataset collected in campus Skolkovo Institute Science Technology (Skoltech). For training, use ORB features that...

Existing visual SLAM approaches are sensitive to illumination, with their precision drastically falling in dark conditions due feature extractor limitations. The algorithms currently used overcome this issue not able provide reliable results poor performance and noisiness, the localization quality is still insufficient for practical use. In paper, we present a novel method capable of working low light using Generative Adversarial Network (GAN) preprocessing module enhance on input images,...

Motion planning in dynamically changing environments is one of the most complex challenges autonomous driving. Safety a crucial requirement, along with driving comfort and speed limits. While classical sampling-based, lattice-based, optimization-based methods can generate smooth short paths, they often do not consider dynamics environment. Some techniques it, but rely on updating environment on-the-go rather than explicitly accounting for dynamics, which suitable self-driving. To address...

In this study, we propose a novel visual localization approach to accurately estimate six degrees of freedom (6-DoF) poses the robot within 3D LiDAR map based on data from an RGB camera. The is obtained utilizing advanced LiDAR-based simultaneous and mapping (SLAM) algorithm capable collecting precise sparse map. features extracted camera images are compared with points map, then geometric optimization problem being solved achieve localization. Our allows employing scout equipped expensive...

SLAM is one of the most fundamental areas research in robotics and computer vision. State art solutions has advanced significantly terms accuracy stability. Unfortunately, not all approaches are available as open-source free to use. The results some them difficult reproduce, there a lack comparison on common datasets. In our work, we make comparative analysis state methods. We assess algorithms based accuracy, computational performance, robustness, fault tolerance. Moreover, present datasets...

In this paper, we propose a novel approach to wheeled robot navigation through an environment with movable obstacles. A exploits knowledge about different obstacle classes and selects the minimally invasive action perform clear path. We trained convolutional neural network (CNN), so can classify RGB-D image decide whether push blocking object which force apply. After known objects are segmented, they being projected cost-map, calculates optimal path goal. If allowed be moved, drives them...

Visual localization is a critical task in mobile robotics, and researchers are continuously developing new approaches to enhance its efficiency. In this article, we propose novel approach improve the accuracy of visual using Structure from Motion (SfM) techniques. We highlight limitations global SfM, which suffers high latency, challenges local requires large image databases for accurate reconstruction. To address these issues, utilizing Neural Radiance Fields (NeRF), as opposed databases,...

Visual localization is an essential modern technology for robotics and computer vision. Popular approaches solving this task are image-based methods. Nowadays, these methods have low accuracy a long training time. The reasons the lack of rigid-body projective geometry awareness, landmark symmetry, homogeneous error assumption. We propose heterogeneous loss function based on concentrated Gaussian distribution with Lie group to overcome difficulties. Following our experiment, proposed method...

Visual localization is a fundamental task for wide range of applications in the field robotics. Yet, it still complex problem with no universal solution, and existing approaches are difficult to scale: most state-of-the-art solutions unable provide accurate without significant amount storage space. We propose hierarchical, low-memory approach based on keypoints different descriptor lengths. It becomes possible use developed unsupervised neural network, which predicts feature pyramid lengths...

Motion planning in dynamically changing environments is one of the most complex challenges autonomous driving. Safety a crucial requirement, along with driving comfort and speed limits. While classical sampling-based, lattice-based, optimization-based methods can generate smooth short paths, they often do not consider dynamics environment. Some techniques it, but rely on updating environment on-the-go rather than explicitly accounting for dynamics, which suitable self-driving. To address...

Visual localization is a fundamental task for wide range of applications in the field robotics. Yet, it still complex problem with no universal solution, and existing approaches are difficult to scale: most state-of-the-art solutions unable provide accurate without significant amount storage space. We propose hierarchical, low-memory approach based on keypoints different descriptor lengths. It becomes possible use developed unsupervised neural network, which predicts feature pyramid lengths...

Visual localization is a critical task in mobile robotics, and researchers are continuously developing new approaches to enhance its efficiency. In this article, we propose novel approach improve the accuracy of visual using Structure from Motion (SfM) techniques. We highlight limitations global SfM, which suffers high latency, challenges local requires large image databases for accurate reconstruction. To address these issues, utilizing Neural Radiance Fields (NeRF), as opposed databases,...

In this paper, we propose a novel approach to wheeled robot navigation through an environment with movable obstacles. A exploits knowledge about different obstacle classes and selects the minimally invasive action perform clear path. We trained convolutional neural network (CNN), so can classify RGB-D image decide whether push blocking object which force apply. After known objects are segmented, they being projected cost-map, calculates optimal path goal. If allowed be moved, drives them...

Existing visual SLAM approaches are sensitive to illumination, with their precision drastically falling in dark conditions due feature extractor limitations. The algorithms currently used overcome this issue not able provide reliable results poor performance and noisiness, the localization quality is still insufficient for practical use. In paper, we present a novel method capable of working low light using Generative Adversarial Network (GAN) preprocessing module enhance on input images,...

Existing Simultaneous Localization and Mapping (SLAM) approaches are limited in their scalability due to growing map size long-term robot operation. Moreover, processing such maps for localization planning tasks leads the increased computational resources required onboard. To address problem of memory consumption operation, we develop a novel real-time SLAM algorithm, MeSLAM, that is based on neural field implicit representation. It combines proposed global mapping strategy, including...

In this study, we propose a novel visual localization approach to accurately estimate six degrees of freedom (6-DoF) poses the robot within 3D LiDAR map based on data from an RGB camera. The is obtained utilizing advanced LiDAR-based simultaneous and mapping (SLAM) algorithm capable collecting precise sparse map. features extracted camera images are compared with points map, then geometric optimization problem being solved achieve localization. Our allows employing scout equipped expensive...

In the proposed study, we describe an approach to improving computational efficiency and robustness of visual SLAM algorithms on mobile robots with multiple cameras limited power by implementing intermediate layer between pipeline. this layer, images are classified using a ResNet18-based neural network regarding their applicability robot localization. The is trained six-camera dataset collected in campus Skolkovo Institute Science Technology (Skoltech). For training, use ORB features that...