检索结果-内蒙古大学图书馆

Evaluating adversarial attacks on driving safety in vision-based autonomous vehicles

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhang, Jindi Lou, Yang Wang, Jianping Wu, Kui Lu, Kejie Jia, Xiaohua The Department of Computer Science City University of Hong Kong Hong Kong The Department of Computer Science University of Victoria VictoriaBCV8P 5C2 Canada The Department of Computer Science and Engineering University of Puerto Rico at Mayagüez Mayagüez00682 Puerto Rico

In recent years, many deep learning models have been adopted in autonomous driving. At the same time, these models introduce new vulnerabilities that may compromise the safety of autonomous vehicles. Specifically, recent studies have demonstrated that adversarial attacks can cause a significant decline in detection precision of deep learning-based 3D object detection models. Although driving safety is the ultimate concern for autonomous driving, there is no comprehensive study on the linkage between the performance of deep learning models and the driving safety of autonomous vehicles under adversarial attacks. In this paper, we investigate the impact of two primary types of adversarial attacks, perturbation attacks and patch attacks, on the driving safety of vision-based autonomous vehicles rather than the detection precision of deep learning models. In particular, we consider two state-of-the-art models in vision-based 3D object detection, Stereo R-CNN and DSGN. To evaluate driving safety, we propose an end-to-end evaluation framework with a set of driving safety performance metrics. By analyzing the results of our extensive evaluation experiments, we find that (1) the attack's impact on the driving safety of autonomous vehicles and the attack's impact on the precision of 3D object detectors are decoupled, and (2) the DSGN model demonstrates stronger robustness to adversarial attacks than the Stereo R-CNN model. In addition, we further investigate the causes behind the two findings with an ablation study. The findings of this paper provide a new perspective to evaluate adversarial attacks and guide the selection of deep learning models in autonomous driving. Copyright © 2021, The Authors. All rights reserved.

关键词： Autonomous vehicles

SurgicalGS: Dynamic 3D Gaussian Splatting for Accurate Robotic-Assisted Surgical Scene Reconstruction

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Chen, Jialei Zhang, Xin Islam, Mobarakol Vasconcelos, Francisco Stoyanov, Danail Elson, Daniel S. Huang, Baoru The Hamlyn Centre for Robotic Surgery Imperial College London SW7 2AZ United Kingdom Department of Mechanical and Aerospace Engineering The HongKong University of Science and Technology Hong Kong Hawkes Institute University College London WC1E 6BT United Kingdom Department of Computer Science University of Liverpool L69 7ZX United Kingdom

Accurate 3D reconstruction of dynamic surgical scenes from endoscopic video is essential for robotic-assisted surgery. While recent 3D Gaussian Splatting methods have shown promise in achieving high-quality reconstructions with fast rendering speeds, their use of inverse depth loss functions compresses depth variations. This can lead to a loss of fine geometric details, limiting their ability to capture precise 3D geometry and effectiveness in intraoperative application. To address these challenges, we present SurgicalGS, a dynamic 3D Gaussian Splatting framework specifically designed for surgical scene reconstruction with improved geometric accuracy. Our approach first initialises a Gaussian point cloud using depth priors, employing binary motion masks to identify pixels with significant depth variations and fusing point clouds from depth maps across frames for initialisation. We use the Flexible Deformation Model to represent dynamic scene and introduce a normalised depth regularisation loss along with an unsupervised depth smoothness constraint to ensure more accurate geometric reconstruction. Extensive experiments on two real surgical datasets demonstrate that SurgicalGS achieves state-of-the-art reconstruction quality, especially in terms of accurate geometry, advancing the usability of 3D Gaussian Splatting in robotic-assisted surgery. © 2024, CC BY.

关键词： 3D reconstruction

You'll Never Walk Alone: Human-Robot Collaborative Navigation in Virtual Environments for Crowd Management

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Chen, Xing Zhao, Hantao Chen, Mashenyi Schläfli, Fabian Thrash, Tyler Cao, Jiuxin Liu, Bo Hölscher, Christoph Shi, Xiaomeng School of Cyber Science and Engineering Southeast University Southeast University Road 2 Nanjing211189 China Purple Mountain Laboratories Nanjing211111 China ETH Zürich Clausiusstrasse 59 Zürich8092 Switzerland Department of Biology Saint Louis University 1 N Grand Blvd St. LouisMO63103 United States School of Computer Science and Engineering Southeast University Southeast University Road 2 Nanjing211189 China School of Transportation Southeast University Southeast University Road 2 Nanjing211189 China

Crowd management in public spaces is a critical challenge faced by modern urban governance, which needs to delicately address the balance between ensuring public safety and respecting individual rights and freedoms. The development of artificial intelligence and robotics technology has provided new avenues for tackling this challenge. This paper proposes a human-robot collaborative navigation model to optimize the efficiency of robot-assisted navigation and enhance the safety of the crowd management process. The model can dynamically adjust the robot's guiding behavior through a state transition mechanism and employ a collaborative scheduling algorithm to control the crowd density in key areas. Additionally, it utilizes dynamic instruction system to provide pedestrians with directional guidance and behavioral recommendations, thereby enhancing the effectiveness of human-robot collaborative navigation. The model's efficacy is validated through empirical research in a virtual reality scenario. The results demonstrate that the human-robot collaborative navigation model has a significant impact on improving navigation efficiency, controlling crowd flow, ensuring traffic safety, and enhancing user satisfaction. © 2024, The Authors. All rights reserved.

关键词： Efficiency

Values of Games for Information Decomposition

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Kroupa, Tomáš Vannucci, Sara Votroubek, Tomáš Artificial Intelligence Center Department of Computer Science Faculty of Electrical Engineering Czech Technical University in Prague Karlovo náměstí 13 Praha 212O OO Czech Republic

The information decomposition problem requires an additive decomposition of the mutual information between the input and target variables into nonnegative terms. The recently introduced solution to this problem, Information Attribution, involves the Shapley-style value measuring the influence of predictors in the coalitional game associated with the joint probability distribution of the input random vector and the target variable. Motivated by the original problem, we consider a general setting of coalitional games where the players form a boolean algebra, and the coalitions are the corresponding down-sets. This enables us to study in detail various single-valued solution concepts, called values. Namely, we focus on the classes of values that can represent very general alternatives to the solution of the information decomposition problem, such as random-order values or sharing values. We extend the axiomatic characterization of some classes of values that were known only for the standard coalitional *** Codes 94A15, 94A17, 91A12, 91A80 Copyright © 2022, The Authors. All rights reserved.

关键词： Boolean algebra

Introduction of the Study Results of the Moon Electrical Power Systems 22

学校读者我要写书评

暂无评论

Introduction of the Study Results of the Moon Electrical Pow...

22nd IAA Symposium on Building Blocks for Future Space Exploration and Development at the 75th International Astronautical Congress, IAC 2024

作者： Oka, Noriaki Yamauchi, Kenji Ijichi, Koichi Yanagawa, Hiroki Kashimura, Osamu Uchida, Atsushi Tanaka, Koji Mitani, Tomohiko Honjo, Kazuhiko Ishikawa, Ryo Ishimura, Kosei Fujino, Yoshiyuki Fujita, Kazuhisa Itoh, Kenji Kawakami, Yoshihiro Ishikawa, Yoji Yamaguchi, Hiroshi Ogawa, Akihiko Japan Space Systems 3-5-8 Shibakoen Minato-ku Tokyo105-0011 Japan Emerging Technologies and Security Division Mitsubishi Research Institute Inc. 10-3 Nagatacho 2-chome Chiyoda-ku Tokyo100-8141 Japan Institute of Space and Astronautical Science Japan Aerospace Exploration Agency 3-1-1 Yoshinodai Chuo-ku Kanagawa Sagamihara252-5210 Japan Kyoto University Gokasho Kyoto Uji611-0011 Japan Department of Computer and Network Engineering The University of Electro-Communications 1-5-1 Chofugaoka Chofu Tokyo182-8585 Japan Dept. of Modern Mechanical Engineering School of Creative Science and Engineering Waseda University 3-4-1 Okubo Shinjuku-ku Tokyo169-8555 Japan Faculty of Science and Engineering Toyo University 2100 Kujirai Saitama-ken Kawagoe350-8585 Japan 1955-1 Kurematsu-cho chuo-ku Shizuoka Hamamatsu431-1202 Japan Department of Electrical and Electronic Engineering College of Engineering Kanazawa Institute of Technology 7-1 Ohgigaoka Ishikawa Nonoichi921-8501 Japan Obayashi Future Lab. Technology Division Obayashi Corporation 2-15-2 Konan Minato-ku Tokyo108-8502 Japan 492 Minosho-machi Nara-ken YamatoKoriyama639-1186 Japan Electronics Technologies Office Technologies Development Department IHI Aerospace Co. Ltd. 900 Fujiki Gunma-ken Tomioka-shi370-2398 Japan

ISBN: (纸本)9798331312251

Since the year of 2021, Ministry of Economy, Trade and Industry (METI) started work on the energy issues on the Moon including the feasibility study of the electrical power system with Mitsubishi Research Institute, Japan Space Systems, Institute of Space and Astronautical science of Japan Aerospace Exploration Agency (JAXA), the universities and the industries. The architecture of the Moon electrical power infrastructure is found to be more like those of the power infrastructure of the earth than current satellite power system, and consisted of the power generation, power distribution grid and the power storage, and supplying electrical power to the consumer. The differences are that the environment of the Moon such as the vacuum and temperature condition, the influence of the regolith, etc. should have to be considered, and also the transportation cost of the construction is the key factor for the necessary technology to be selected. Of course, the technologies already established by the International Space Station (ISS) should be applied where necessary such as in the human habitable area. And more feature of the Moon is that the power generation on the Moon orbit and be sent to the surface or the inside the Permanent Shadow Region (PSR) other than on the surface of the Moon. The main consumer of the power system until the year 2050th is the water extraction and then the electrolysis of the water for, such as, the propellant for the transportation means and the electricity for the human habitant. The results of the system feasibility study of the Moon power architecture and also the electric power system team activities including the discussion with the hydrogen production team are going to be introduced. Copyright © 2024 by the International Astronautical Federation (IAF). All rights reserved.

关键词： Space power generation

A4T: Hierarchical Affordance Detection for Transparent Objects Depth Reconstruction and Manipulation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Jiang, Jiaqi Cao, Guanqun Do, Thanh-Toan Luo, Shan SmARTLab Department of Computer Science University of Liverpool LiverpoolL69 3BX United Kingdom Department of Engineering King's College London LondonWC2R 2LS United Kingdom Department of Data Science and AI Monash University ClaytonVIC3800 Australia

Transparent objects are widely used in our daily lives and therefore robots need to be able to handle them. However, transparent objects suffer from light reflection and refraction, which makes it challenging to obtain the accurate depth maps required to perform handling tasks. In this paper, we propose a novel affordance-based framework for depth reconstruction and manipulation of transparent objects, named A4T. A hierarchical AffordanceNet is first used to detect the transparent objects and their associated affordances that encode the relative positions of an object's different parts. Then, given the predicted affordance map, a multi-step depth reconstruction method is used to progressively reconstruct the depth maps of transparent objects. Finally, the reconstructed depth maps are employed for the affordance-based manipulation of transparent objects. To evaluate our proposed method, we construct a real-world dataset TRANS-AFF with affordances and depth maps of transparent objects, which is the first of its kind. Extensive experiments show that our proposed methods can predict accurate affordance maps, and significantly improve the depth reconstruction of transparent objects compared to the state-of-the-art method, with the Root Mean Squared Error in meters significantly decreased from 0.097 to 0.042. Furthermore, we demonstrate the effectiveness of our proposed method with a series of robotic manipulation experiments on transparent objects. See supplementary video and results at https://***/view/affordance4trans. Copyright © 2022, The Authors. All rights reserved.

关键词： Light reflection

Supervised Representation Learning From Problem-Solving Behavior in Virtual Reality Manufacturing Environment Via Long Short-Term Memory

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Zhu, Rui Aqlan, Faisal Zhao, Richard Yang, Hui Li, Yifu School of Industrial and Systems Engineering The University of Oklahoma NormanOK73019 United States Department of Industrial Engineering University of Louisville LouisvilleKY40292 United States Department of Computer Science University of Calgary CalgaryABT2N 1N4 Canada Complex Systems Monitoring Modeling and Control Laboratory The Pennsylvania State University University ParkPA16802 United States

The number of manufacturing jobs has been consistently increasing thanks to the rapidly evolving landscape and the newly adopted strategic plan of US manufacturing. However, there have been rising concerns in recent years about the problem-solving skills of engineering students, who will soon become the next generation of the US workforce. Such circumstances call for systematic evaluation and a better understanding of engineering students’ problem-solving. This paper builds a virtual reality (VR)-based manufacturing environment and adopts a data-driven analytical workflow to evaluate engineering students’ problem-solving performance. Specifically, students perform assembly tasks to produce car toys that satisfy the pre-defined requirements within the VR craft-production manufacturing system. We collect in situ eye-tracking data containing spatial-temporal dynamics of eye movements and abstract their latent features via a long short-term memory (LSTM)-based supervised representation learning for problem-solving performance evaluation. This work outperforms the traditional performance metrics-based evaluation by incorporating the nonlinear dynamics summarized from the in situ production process. Experimental results show that learned feature representations provide significantly more distinguishable patterns when categorizing the students’ problem-solving performance than traditional performance metrics. The proposed evaluation workflow has a broader impact on facilitating the problem-solving assessment across a wide range of manufacturing systems and worker training programs. © 2024, The Authors. All rights reserved.

关键词： Eye tracking

Crystalline superconductor-semiconductor Josephson junctions for compact superconducting qubits

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Balgley, Jesse Park, Jinho Chu, Xuanjing Arnault, Ethan G. Gustafsson, Martin V. Ranzani, Leonardo Holbrook, Madisen Watanabe, Kenji Taniguchi, Takashi Perebeinos, Vasili Hone, James Fong, Kin Chung Department of Mechanical Engineering Columbia University New YorkNY10027 United States Department of Applied Physics and Mathematics Columbia University New YorkNY10027 United States Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States RTX BBN Technologies Quantum Engineering and Computing Group CambridgeMA02138 United States Department of Physics Columbia University New YorkNY10027 United States National Institute for Materials Science 1-2-1 Sengen Ibaraki Tsukuba305-0044 Japan Department of Electrical Engineering University at Buffalo The State University of New York BuffaloNY14260 United States

The narrow bandgap of semiconductors allows for thick, uniform Josephson junction barriers, potentially enabling reproducible, stable, and compact superconducting qubits. We study vertically stacked van der Waals Josephson junctions with semiconducting weak links, whose crystalline structures and clean interfaces offer a promising platform for quantum devices. We observe robust Josephson coupling across 2–12 nm (3–18 atomic layers) of semiconducting WSe2 and, notably, a crossover from proximity- to tunneling-type behavior with increasing weak link thickness. Building on these results, we fabricate a prototype all-crystalline merged-element transmon qubit with transmon frequency and anharmonicity closely matching design parameters. We demonstrate dispersive coupling between this transmon and a microwave resonator, highlighting the potential of crystalline superconductor-semiconductor structures for compact, tailored superconducting quantum devices. Copyright © 2025, The Authors. All rights reserved.

关键词： Josephson junction devices

Assessing alternatives of including social robots in urban transport using fuzzy trigonometric operators based decision-making model

学校读者我要写书评

暂无评论

Technological Forecasting and Social Change 2023年 194卷

作者： Deveci, Muhammet Pamucar, Dragan Gokasar, Ilgin Zaidan, Bilal Bahaa Martinez, Luis Pedrycz, Witold Department of Industrial Engineering Turkish Naval Academy National Defence University Istanbul Tuzla34940 Turkey Royal School of Mines Imperial College London LondonSW7 2AZ United Kingdom Department of Operations Research and Statistics Faculty of Organizational Sciences University of Belgrade Belgrade11000 Serbia Department of Civil Engineering Bogazici University Istanbul Bebek34342 Turkey International Graduate School of Artificial Intelligence College of Management National Yunlin University of Science and Technology 123 University Road Section 3 Yunlin Douliu City64002 Taiwan Department of Computer Sciences University of Jaén Jaén23071 Spain Department of Electrical and Computer Engineering Faculty of Engineering University of Alberta 9211 116 Street NW EdmontonABT6G 1H9 Canada Systems Research Institute Polish Academy of Sciences Warsaw00-901 Poland Faculty of Engineering and Natural Sciences Department of Computer Engineering Istinye University Sariyer/Istanbul Turkey College of Engineering Yuan Ze University Taiwan

Current trends point to a not-too-distant future with qualitatively advanced interactions between humans and social robots. It is critical to consider the possibility of forming meaningful social relationships with robots when defining the future of human-robot interactions, as well as studying how these interactions will evolve to the point where humans are unable to distinguish between humans and robots in urban transportation. In this study, the advantages of using social robots in urban transportation are prioritized by using a multi-criteria decision-making tool, which consists of two consecutive stages, namely: i) a novel fuzzy sine trigonometry based on the logarithmic method of additive weights (fuzzy ST-LMAW) that is proposed to calculate the criteria weights;ii) a nonlinear fuzzy Aczel-Alsina function based the weighted aggregate sum product assessment (fuzzy ALWAS-WASPAS) that is developed to select and rank the alternatives. The proposed model enables flexible nonlinear processing of complex and uncertain information encountered in real applications. A case study is developed to rank three alternatives with twelve sub-criteria grouped into four aspects using the proposed method. The results show that the most advantageous alternative is to replace people with social robots as safety drivers in level four autonomous vehicles due to their possible impact on transportation. © 2023 The Authors

关键词： Human robot interaction