检索结果-内蒙古大学图书馆

LiFT: Lightweight, FPGA-tailored 3D object detection based on LiDAR data

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Lis, Konrad Kryjak, Tomasz Gorgoń, Marek Embedded Vision Systems Group Department of Automatic Control and Robotics AGH University of Krakow Al. Mickiewicza 30 Krakow30-059 Poland

This paper presents LiFT, a lightweight, fully quantized 3D object detection algorithm for LiDAR data, optimized for real-time inference on FPGA platforms. Through an in-depth analysis of FPGA-specific limitations, we identify a set of FPGA-induced constraints that shape the algorithm’s design. These include a computational complexity limit of 30 GMACs (billion multiply-accumulate operations), INT8 quantization for weights and activations, 2D cell-based processing instead of 3D voxels, and minimal use of skip connections. To meet these constraints while maximizing performance, LiFT combines novel mechanisms with state-of-the-art techniques such as reparameterizable convolutions and fully sparse architecture. Key innovations include the Dual-bound Pillar Feature Net, which boosts performance without increasing complexity, and an efficient scheme for INT8 quantization of input features. With a computational cost of just 20.73 GMACs, LiFT stands out as one of the few algorithms targeting minimal-complexity 3D object detection. Among comparable methods, LiFT ranks first, achieving an mAP of 51.84% and an NDS of 61.01% on the challenging NuScenes validation dataset. The code will be available at https://***/vision-agh/lift. © 2025, CC BY.

关键词：

Target Defense with Multiple Defenders and an Agile Attacker via Residual Policy Learning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Tao, Jiyue Shen, Tongsheng Zhao, Dexin Zhang, Feitian Robotics and Control Laboratory Department of Advanced Manufacturing and Robotics College of Engineering and the State Key Laboratory of Turbulence and Complex Systems Peking University Beijing100871 China National Innovation Institute of Defense Technology Beijing100071 China

The target defense problem involves intercepting an attacker before it reaches a designated target region using one or more defenders. This letter focuses on a particularly challenging scenario in which the attacker is more agile than the defenders, significantly increasing the difficulty of effective interception. To address this challenge, we propose a novel residual policy framework that integrates deep reinforcement learning (DRL) with the force-based Boids model. In this framework, the Boids model serves as a baseline policy, while DRL learns a residual policy to refine and optimize the defenders’ actions. Simulation experiments demonstrate that the proposed method consistently outperforms traditional interception policies, whether learned via vanilla DRL or fine-tuned from force-based methods. Moreover, the learned policy exhibits strong scalability and adaptability, effectively handling scenarios with varying numbers of defenders and attackers with different agility levels. Copyright © 2025, The Authors. All rights reserved.

关键词： Adversarial machine learning

Vision-based automatic fruit counting with UAV

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Szolc, Hubert Wasala, Mateusz Mietla, Remigiusz Iwicki, Kacper Kryjak, Tomasz Embedded Vision Systems Group Computer Vision Laboratory Department of Automatic Control and Robotics AGH University of Krakow al. Mickiewicza 30 Krakow30–059 Poland

The use of unmanned aerial vehicles (UAVs) for smart agriculture is becoming increasingly popular. This is evidenced by recent scientific works, as well as the various competitions organised on this topic. Therefore, in this work we present a system for automatic fruit counting using UAVs. To detect them, our solution uses a vision algorithm that processes streams from an RGB camera and a depth sensor using classical image operations. Our system also allows the planning and execution of flight trajectories, taking into account the minimisation of flight time and distance covered. We tested the proposed solution in simulation and obtained an average score of 87.27/100 points from a total of 500 missions. We also submitted it to the UAV Competition organised as part of the ICUAS 2024 conference, where we achieved an average score of 84.83/100 points, placing 6th in a field of 23 teams and advancing to the finals. © 2025, CC BY.

关键词：

Enhancing Thin-Film Wafer Inspection With A Multi-Sensor Array And Robot Constraint Maintenance

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Sánchez-Arriaga, Néstor Eduardo Canzini, Ethan Espley-Plumb, Nathan John Farnsworth, Michael Pope, Simon Leyland, Adrian Tiwari, Ashutosh Department of Automatic Control & Systems Engineering University of Sheffield Sheffield United Kingdom Department of Materials Science & Engineering University of Sheffield Sheffield United Kingdom Airbus Robotics Airbus UK Chester United Kingdom

Thin-film inspection on large-area substrates in coating manufacture remains a critical parameter to ensure product quality;however, extending the inspection process precisely over a large area presents major challenges, due to the limitations of the available inspection equipment. An additional manipulation problem arises when automating the inspection process, as the silicon wafer requires movement constraints to ensure accurate measurements and to prevent damage. Furthermore, there are other increasingly important large-area industrial applications, such as Roll-to-Roll (R2R) manufacturing where coating thickness inspection introduces additional challenges. This paper presents an autonomous inspection system using a robotic manipulator with a novel learned constraint manifold to control a wafer to its calibration point, and a novel multi-sensor array with high potential for scalability into large substrate areas. We demonstrate that the manipulator can perform required motions whilst adhering to movement constraints. We further demonstrate that the sensor array can perform thickness measurements statically with an error of 0.5◦ from the calibration point whilst monitoring the RMSE and R2 over 1406 data points. These features are potentially useful for detecting displacement variations in R2R manufacturing processes. © 2025, CC BY-NC-ND.

关键词： Silicon wafers

Retinal disorder diagnosis based on hybrid deep learning models

学校读者我要写书评

暂无评论

Multimedia Tools and Applications 2025年 1-25页

作者： Sedik, Ahmed El-Shafai, Walid El-Hag, Noha A. El-Banby, Ghada M. Abd El-Samie, Fathi E. Smart Systems Engineering Laboratory College of Engineering Prince Sultan University Riyadh11586 Saudi Arabia Department of the Robotics and Intelligent Machines Faculty of Artificial Intelligence Kafrelsheikh University Kafr el-Sheikh Egypt Computer Science Department Prince Sultan University Riyadh11586 Saudi Arabia Department of Electronics and Electrical Communications Engineering Faculty of Electronic Engineering Menoufia University Menouf32952 Egypt The Higher Institute of Commercial Sciences Algarbia Al Mahalla Al Kubra31951 Egypt Department of Industrial Electronics and Control Engineering Faculty of Electronic Engineering Menoufia University Menouf32952 Egypt Department of Information Technology College of Computer and Information Sciences Princess Nourah Bint Abdulrahman University P.O. Box 84428 Riyadh11671 Saudi Arabia

Anomaly detection from medical images is badly needed for automated diagnosis. For example, medical images obtained with several modalities, such as magnetic resonance (MR) and confocal microscopy, need to be classified for anomaly detection. The MR images may contain tumors that need to be detected with anomaly detection techniques. In MR imaging, brain tumors are represented as areas composed of abnormal cells that need to be detected. On the other hand, confocal microscopy can be used for retinal imaging. Hence, there is a need to apply classification techniques for diagnosing diseases of the eye, such as Diabetic Retinopathy (DR), which leads to exudates in different regions of the retina. The detection of exudates is necessary for the diagnosis of DR. In this paper, two models are presented for the classification of MR and retinal images. Much research has been done in this area, but it did not achieve high accuracy of classification in an acceptable time. In addition, the previous models did not consider both types of image modalities considered in this paper. The suggested models are Convolutional Neural Network (CNN) and Convolutional Long Short-Term Memory (Conv-LSTM). They are used to classify normal and abnormal cases of both brain tumor and DR diseases. The CNN model is composed of six convolutional layers and six max-pooling layers. A soft-max activation function takes the classification decision. The proposed Conv-LSTM model is composed of three stages: pre-processing, feature extraction, and classification. The deep learning model performance is better than those of traditional networks. The CNN gives an accuracy of 94.26% for DR classification and 95.8% for brain tumor classification, while the Conv-LSTM gives the maximum accuracy of 96.41% for DR classification and 97.4% for brain tumor classification. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2025.

关键词： Confocal microscopy

Dynamic Collision Avoidance Using Velocity Obstacle-Based control Barrier Functions

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Huang, Jihao Zeng, Jun Chi, Xuemin Sreenath, Koushil Liu, Zhitao Su, Hongye State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China Institute of Intelligence Science and Engineering Shenzhen Polytechnic University Shenzhen China Hybrid Robotics Group The Department of Mechanical Engineering UC Berkeley United States

Designing safety-critical controllers for acceleration-controlled unicycle robots is challenging, as control inputs may not appear in the constraints of control Lyapunov functions (CLFs) and control barrier functions (CBFs), leading to invalid controllers. Existing methods often rely on state-feedback-based CLFs and high-order CBFs (HOCBFs), which are computationally expensive to construct and fail to maintain effectiveness in dynamic environments with fast-moving, nearby obstacles. To address these challenges, we propose constructing velocity obstacle-based CBFs (VOCBFs) in the velocity space to enhance dynamic collision avoidance capabilities, instead of relying on distance-based CBFs that require the introduction of HOCBFs. Additionally, by extending VOCBFs using variants of VO, we enable reactive collision avoidance between robots. We formulate a safety-critical controller for acceleration-controlled unicycle robots as a mixed-integer quadratic programming (MIQP), integrating state-feedback-based CLFs for navigation and VOCBFs for collision avoidance. To enhance the efficiency of solving the MIQP, we split the MIQP into multiple sub-optimization problems and employ a decision network to reduce computational costs. Numerical simulations demonstrate that our approach effectively guides the robot to its target while avoiding collisions. Compared to HOCBFs, VOCBFs exhibit significantly improved dynamic obstacle avoidance performance, especially when obstacles are fast-moving and close to the robot. Furthermore, we extend our method to distributed multi-robot systems. Copyright © 2025, The Authors. All rights reserved.

关键词： State feedback

Machine Learning in Predicting the Impact of Filter Parameters on the Accuracy of Biosignal Measurement

学校读者我要写书评

暂无评论

Machine Learning in Predicting the Impact of Filter Paramete...

Digital Signal Processing and its Applications (DSPA), International Conference on

作者： Yeldos A. Altay Raisa K. Uskenbayeva Abu-Alim E. Ayazbay Alexey V. Fedorov Department of Robotics and Technical Means of Automation Satbayev University Almaty Kazakhstan Institute of Automation and Information Technologies Vice-Rector Satbayev University Almaty Kazakhstan Electronic engineering Almaty University of Power Engineering and Telecomunications Almaty Kazakhstan Faculty of Control Systems and Robotics ITMO University Saint-Petersburg Russian Federation

ISBN: (数字)9798331518752

ISBN: (纸本)9798331518769

In this scientific article under consideration, a machine learning model has been developed that allows predicting the impact of filter parameters on the accuracy of measuring a biosignal - an ECG (electrocardiography) signal. For the first time, quantitative results have shown the degree of influence and the error introduced into the measured biosignal and described through dependence. The results of assessing the obtained relationship of indicators demonstrate an inverse correlation relationship that characterizes a decrease in the accuracy of measurement at the output of a high-order digital filtering system. The machine learning model is based on the linear regression method the dependence of which is approximated by the method based on least squares and visualized using a scattergram. The adequacy of the developed machine learning model was confirmed by calculating the coefficient of determination and checking its statistical significance. The developed model can be used to create new or improve existing biomedical data processing systems to increase the reliability of the results.

关键词： Accuracy Shape Filtering Biological system modeling Machine learning Biomedical measurement Electrocardiography Distortion Distortion measurement Digital filters

Human-Robot Pose Tracking Based on CNN with Color and Geometry Aggregation 16th

学校读者我要写书评

暂无评论

Human-Robot Pose Tracking Based on CNN with Color and Geo...

16th International Conference on Social robotics, ICSR + InnoBiz 2024

作者： Xu, Yue Zhang, Yinlong Liu, Shuai Liu, Yuanhao Liang, Wei He, Hongsheng Shenyang University of Technology Shenyang110020 China Guangzhou Institute of Industrial Intelligence Guangzhou511458 China State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Sciences Shenyang110016 China Key Laboratory of Networked Control Systems Chinese Academy of Sciences Shenyang110016 China University of Chinese Academy of Sciences Beijing100049 China Department of Computer Science The University of Alabama TuscaloosaAL35487 United States

ISBN: (纸本)9789819611508

Accurately tracking the robotic arm and human joints is crucial to ensure safety during human-robot interaction. However, traditional pose tracking methods often exhibit insufficient performance and robustness in complex environments. The variations in the robotic arm’s environment and posture make it challenging for traditional methods to accurately capture the positions and posture of its joints. Specifically, when addressing challenges such as high similarity, occlusion, background complexity, and joint recognition failures, they often struggle to provide reliable and accurate results. To address these challenges, this paper proposes a human-robot pose tracking algorithm based on a new convolutional neural network model. To enhance detection accuracy, an improved color detection module is introduced to resolve joint misclassification. A geometric perception module is designed to accurately locate joints even under occlusion. Additionally, innovative iEMA and DBB modules are incorporated. The iEMA module employs edge detection technology to dynamically adjust thresholds for correct matching by improving the edge matching process. The DBB module refines the boundary box parameters for precise localization by introducing adaptive bounding box updates in real-time. This algorithm also integrates human pose recognition, enabling real-time pose recognition, thereby facilitating more intelligent and natural human-robot interaction. The algorithm has been rigorously evaluated on a custom-designed robotic arm platform. Experimental results validate the algorithm’s effectiveness and feasibility. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Robotic arms

Electric Vehicle Routing Problem with Capacitated Charging Stations and Dynamic Electricity Prices

学校读者我要写书评

暂无评论

IEEE Internet of Things Journal 2025年

作者： He, Zhiye Tian, Yuning Miao, Haoyu Yao, Canqi Yang, Zaiyue Shenzhen Key Laboratory of Control Theory and Intelligent Systems Guangdong Provincial Key Laboratory of Fully Actuated System Control Theory and Technology Guangdong Provincial Key Laboratory of Human-Augmentation and Rehabilitation Robotics in Universities School of System Design and Intelligent Manufacturing Shenzhen518055 China The Hong Kong Polytechnic University Department of Logistics and Maritime Studies Hong Kong Hong Kong

With rapid electrification of transportation systems in recent years, the electric vehicle routing problem (EVRP) has received increasing attention. In order to serve a set of valuable transportation tasks using a fleet of electric vehicles, EVRP intends to jointly determine their routes and the charging schedules, with the aim of optimizing service reward and charging cost under state of charge (SOC) constraints of battery. EVRP is a difficult nonlinear integer program, and due to its intrinsic NP-hardness early EVRP studies often oversimplify the charging process and assume unlimited capacity of charging stations and constant electricity price. The contribution of this paper lies in twofold. First, to better characterize the practical charging process, this paper introduces a new graph model that can take the charging station capacity and dynamic electricity price into full consideration. Second, we design a series of transformation techniques, including reformulation SOC dynamics, Lagrangian relaxation, exact linear programming relaxation, etc, which eventually convert the original EVRP into an approximate linear program. Extensive simulations across 14 instances demonstrate that, the proposed approach is capable of solving large-scale instances over 500 nodes, and is 100 times faster than traditional analytic methods while achieving very close optimality. While comparing with the popular heuristic method ALNS (Adaptive Large Neighborhood Search), our method is several times faster and obtains much better optimality. © 2014 IEEE.

关键词： Charging stations