Remote sensing image scene classification (RSI-SC) is crucial for various high-level applications, including RSI retrieval, captioning, and object detection. Deep learning-based methods can accurately predict categories. However, these approaches often require numerous labeled samples training, limiting their practicality in real-world RS applications with scarce label resources. In contrast, few-shot remote (FS-RSI-SC) has garnered substantial research interest owing to its potential...

This paper proposes an efficient and robust Loop closure detection (LCD) method based on Convolutional neural network (CNN) feature. The primary is called SeqCNNSLAM, in which both the outputs of intermediate layer a pre-trained CNN traditional sequence-based matching procedure are incorporated, making it possible to handle viewpoint condition variance properly. An acceleration algorithm for SeqCNNSLAM developed reduce search range current image, resulting new LCD A-SeqCNNSLAM. To improve...

Vision Transformer (ViT) models have recently emerged as powerful and versatile tools for various visual tasks. In this article, we investigate ViT in a more challenging scenario within the context of few-shot conditions. Recent work has achieved promising results image classification by utilizing pre-trained vision transformer models. However, employs full fine-tuning downstream tasks, leading to significant overfitting storage issues, especially remote sensing domain. order tackle these...

In this paper, we present a solution to 3D mapping using 2D laser scanner with pose estimates from an IMU-aided visual SLAM system. Accurate motion estimation of robot is achieved by visual-inertial fusion based on extended Kalman filter (EKF). Range measurements scanned the vertical plane are received constantly mounted robot, which re-organized as point clouds in Cartesian space. With can be transformed into world frame real time. Furthermore, these between two consecutive keyframes system...

The low-cost Inertial Measurement Unit (IMU) can provide orientation information and is widely used in our daily life. However, IMUs with bad calibration will inaccurate angular velocity lead to rapid drift of integral a short time. In this paper, we present the Calib-Net which achieve accurate IMU via simple deep convolutional neural network. Following carefully designed mathematical model, output compensation components for gyroscope measurements dynamically. Dilation convolution adopted...

In this paper, we consider a surveillance scenario where team of sensing robots survey sensitive area and transmit the monitored data to remote base station through mobile relay. scenario, it is challenging autonomously adjust position relay for sake minimizing total communication-motion energy consumption system, while maintaining communication quality robots. We first derive asymptotically optimal powers according predefined end-to-end packet error rate (PER) requirement. Then, propose...

This letter focuses on a scenario in which team of sensing robots survey an area with predefined routes, and transmit the monitored information to remote base station through mobile relay. In this scenario, automatically adjusting position relay for maintaining wireless link quality while are moving is challenging problem. letter, we consider problem minimizing total energy consumption. We propose using dynamic programming (DP) single-step optimization. By comparing pros cons both methods,...

This paper presents an innovative exploration of the application potential large language models (LLM) in addressing challenging task automatically generating behavior trees (BTs) for complex tasks. The conventional manual BT generation method is inefficient and heavily reliant on domain expertise. On other hand, existing automatic technologies encounter bottlenecks related to complexity, model adaptability, reliability. In order overcome these challenges, we propose a novel methodology that...

Programming control systems for mobile robots is complicated and time-consuming, due to three aspects, i.e., the robot behavior coordination, distributed multi-robot cooperation software reusability. Subsumption model a robust architecture robots. ALLIANCE extends it multirobot systems, which fully distributed, fault-tolerant model. Robot operating system (ROS) provides lot of reusable modules. By combining above three, we propose framework named ALLIANCE-ROS developing cooperative with...

For efficient path planning of ground robots in 3D environments with structures such as buildings or overhanging objects, an appropriate spatial representation the environment is normally required. Some popular representations, elevation maps and multi-level surface maps, need to be projected into a 2D plane extract traversibility for planning. They cannot properly handle all complex situations, bridges. other predominant occupancy grid normal distributions typically have high computational...

Purpose – The purpose of this paper is to design intelligent robots operating in such dynamic environments like the RoboCup Middle-Size League (MSL). In MSL, two teams five autonomous play on an 18- × 12-m field. Equipped with sensors and on-board computers, each robot should be able perceive environment, make decision control itself soccer game autonomously. Design/methodology/approach This presents our robots, participating MSL. mechanical platform, electrical architecture software...

Wireless communications and networking are playing an important role in coordination cooperation of multi-robot systems (MRS). However, it is challenging to keep a reliable stable wireless connection practical applications. Especially, robots acting electromagnetic adversarial (EA) environments may encounter more serious situations including scarce spectrum, active interference, competition, etc. In this survey, we firstly analyze the challenges faced by MRS EA environments, provide...

The loop closure detection (LCD) is an essential part of visual simultaneous localization and mapping systems (SLAM). LCD capable identifying compensating the accumulation drift algorithms to produce consistent map if loops are checked correctly. Deep convolutional neural networks (CNNs) have outperformed state-of-the-art solutions that use traditional hand-crafted features in many computer vision pattern recognition applications. After great success CNNs, there has been much interest...

Most state-of-the-art visual simultaneous localization and mapping (SLAM) systems are designed for applications in static environments. However, during a SLAM process, dynamic objects the field-of-view of camera will affect accuracy odometry loop-closure detection. In this paper, we present solution to removing from RGB images their corresponding depth when RGB-D is mounted on mobile robot SLAM. We transform two selected successive same image coordinate frame through feature matching. Then...

In this paper, we present a solution to visual simultaneous localization and mapping (SLAM) using multiple RGB-D cameras. the SLAM system, integrate depth measurements from those cameras achieve more robust pose tracking detailed environmental in unknown environments. We mathematical analysis of iterative optimizations for map refinement system multi-camera cases. The resulted allows configurations with non-overlapping fields view (FOVs). Furthermore, provide SLAM-based semiautomatic method...

Multi-robot systems are widely applied in various areas. The connectivity within robots plays an important role many cooperative tasks. In this paper, we propose a deep Q-network based learning approach for preservation problem. We adopt fully connected neural network with nonlinearities as the Q-function framework. Further, design and implement simulation environment preservation, which provides states reward feedbacks to Q-network. demonstrate framework by series of experiments....

Robots are developing in much the same way that personal computers did 40 years ago, and robot operating system is critical basis. Current software mainly designed for individual robots. We present this paper design of micROS, a morphable, intelligent collective future collaborative first architecture including distributed as whole layered every single node. then autonomous behavior management based on observe–orient–decide–act cognitive model intelligence perception, cognition, game...

The problem of one-on-one target tracking from a single monocular image acquired the viewpoint follower robot itself is studied in this paper. Previous works mainly depended on locating, onboard sensors with control mechanism, while may not carry advanced equipment for localization or GNSS also fail GNSS-denied/Indoor environments. In paper we propose novel approach based deep convolutional neural network called Deep-Track, which trains supervised classifier only using images captured by...

The one-on-one target tracking problem is important in robot vision. Previous studies mainly focused on locating, depth information and control mechanism. In this study, we construct an autonomously visual system called learn-to-track (LtT) by using a novel approach. This only depends monocular camera. main component deep convolutional neural network the LtT, which trains supervised image classifier images captured camera follower robot. By operating merely two adjacent frames, can predict...