检索结果-内蒙古大学图书馆

Efficient Large Scale Stereo Matching based on Cross-Scale

学校读者我要写书评

暂无评论

Efficient Large Scale Stereo Matching based on Cross-Scale

International Conference on control, Automation, Robotics and Vision (ICARCV)

作者： Yue Xia Zhitao Liu Hongye Su Hao Shu Polytechnic Institute Zhejiang University Hangzhou China State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China

ISBN: (纸本)9781665476881

We propose a binocular stereo matching algorithm CS-ELAS. This is a cross-scale ELAS algorithm that improves the accuracy and robustness of parallax in weakly textured regions and edge regions. Our approach focuses on improving the accuracy and number of support point sets. We uniformly sample the stereo images to obtain candidate support point sets and determine robustly matched support point sets based on an adaptive cross skeleton. In this way a richer and more accurate set of support points can be obtained in the weakly textured areas near the edges. In addition, our method uses the parallax and confidence maps of low-resolution images as a priori for high-resolution images and adds high-confidence pixel information to the set of high-resolution support points. In this way, it not only increases the number of support points in weak texture regions, but also narrows the search range of candidate support points and reduces the computational cost.

关键词： Three-dimensional displays Image resolution Automation Image edge detection Benchmark testing Skeleton Robustness

A General Petri-Net controller for Avoiding Collisions and Deadlocks in a Partially-controllable Mixed Traffic System

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Chen, Xinyu Wu, Weimin Hu, Ruifen State Key Laboratory of Industrial Control Technology Zhejiang University Hangzhou310058 China Institute of Cyber-Systems and Control Zhejiang University Hangzhou310058 China

Mixed traffic systems are extensively employed to carry out transportation tasks in a flexible manufacturing system (FMS), especially since automated guided vehicles (AGVs) have become more and more popularized in recent years. However, current researches mainly focus on the coordination of a fully autonomous multi-AGV system, while neglecting the influence of human-driven vehicles in a mixed traffic environment. This paper proposes a Petri-net controller to avoid collisions and deadlocks in a partially-controllable mixed traffic system. As for the theoretical framework of this paper, the mixed traffic system is first modeled by a colored labeled Petri net, where different vehicles are represented by different colors and transitions labeled by an uncontrollable vehicle will be fired randomly. Based on the Petri net structure, the collision-free and deadlock-free conditions are formalized, from which a general controller that avoids collisions and deadlocks in a partially-controllable mixed traffic system is designed. Experiments are carried out to verify the effectiveness of the proposed controller and the results are highly promising in terms of coordinating a mixed traffic system. © 2022, The Authors. All rights reserved.

关键词： Petri nets

Adaptive Robust Tracking control for First-Order Linear systems with Input Delay and Lipschitz Nonlinear Disturbance

学校读者我要写书评

暂无评论

Adaptive Robust Tracking Control for First-Order Linear Syst...

American control Conference (ACC)

作者： Han Lai Yang Zhu Zheng Chen Bin Yao State Key Laboratory of Fluid Power and Mechatronic Systems Zhejiang University Hangzhou China State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China Ocean College Zhejiang University Zhoushan China School of Mechanical Engineering Purdue University West Lafayette IN USA

In this paper, an adaptive robust tracking controller is proposed for first-order linear systems with input delay, unknown plant parameters and Lipschitz nonlinear disturbance. The controller employs the predictor feedback to compensate for the effect of input delay, the robust feedback to deal with uncertainties, the model compensation for trajectory tracking, and projection-type adaptation laws are designed. By the stability analysis with a Lyapunov function in integral form, the closed-loop system is locally stable in the sense that the tracking error is bounded above by a known function which exponentially converges to a specified accuracy provided that the initial states and control parameters meet certain conditions. Furthermore, when the disturbance is reduced to a constant, the controller guarantees the semi-global stability that the tracking error asymptotically converges to zero. Simulation results demonstrate the effectiveness of the proposed controller.

关键词：

DeepRING: Learning Roto-translation Invariant Representation for LiDAR based Place Recognition

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Lu, Sha Xu, Xuecheng Tang, Li Xiong, Rong Wang, Yue State Key Laboratory of Industrial Control and Technology The Institute of Cyber-Systems and Control Zhejiang University Hangzhou310058 China Alibaba Group Hangzhou310052 China

LiDAR based place recognition is popular for loop closure detection and re-localization. In recent years, deep learning brings improvements to place recognition by learnable feature extraction. However, these methods degenerate when the robot re-visits previous places with large perspective difference. To address the challenge, we propose DeepRING to learn the roto-translation invariant representation from LiDAR scan, so that robot visits the same place with different perspective can have similar representations. There are two keys in DeepRING: the feature is extracted from sinogram, and the feature is aggregated by magnitude spectrum. The two steps keeps the final representation with both discrimination and roto-translation invariance. Moreover, we state the place recognition as a one-shot learning problem with each place being a class, leveraging relation learning to build representation similarity. Substantial experiments are carried out on public datasets, validating the effectiveness of each proposed component, and showing that DeepRING outperforms the comparative methods, especially in dataset level generalization. Copyright © 2022, The Authors. All rights reserved.

关键词： Deep learning

Memory-Efficient Reversible Spiking Neural Networks

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhang, Hong Zhang, Yu State Key Laboratory of Industrial Control Technology College of Control Science and Engineering Zhejiang University Hangzhou China Key Laboratory of Collaborative sensing and autonomous unmanned systems of Zhejiang Province Hangzhou China

Spiking neural networks (SNNs) are potential competitors to artificial neural networks (ANNs) due to their high energyefficiency on neuromorphic hardware. However, SNNs are unfolded over simulation time steps during the training process. Thus, SNNs require much more memory than ANNs, which impedes the training of deeper SNN models. In this paper, we propose the reversible spiking neural network to reduce the memory cost of intermediate activations and membrane potentials during training. Firstly, we extend the reversible architecture along temporal dimension and propose the reversible spiking block, which can reconstruct the computational graph and recompute all intermediate variables in forward pass with a reverse process. On this basis, we adopt the state-of-the-art SNN models to the reversible variants, namely reversible spiking ResNet (RevSResNet) and reversible spiking transformer (RevSFormer). Through experiments on static and neuromorphic datasets, we demonstrate that the memory cost per image of our reversible SNNs does not increase with the network depth. On CIFAR10 and CIFAR100 datasets, our RevSResNet37 and RevSFormer-4- 384 achieve comparable accuracies and consume 3.79× and 3.00× lower GPU memory per image than their counterparts with roughly identical model complexity and parameters. We believe that this work can unleash the memory constraints in SNN training and pave the way for training extremely large and deep SNNs. The code is available at https://***/mi804/***. Copyright © 2023, The Authors. All rights reserved.

关键词： Neural networks

SAMOC: Enabling Atomic Invocations for Cross-chain Crowdsourcing Testing DApps in industrial control Through Trusted Smart Community and Lock Mechanism 23

学校读者我要写书评

暂无评论

SAMOC: Enabling Atomic Invocations for Cross-chain Crowdsour...

23rd IEEE International Conference on Trust, Security and Privacy in Computing and Communications, TrustCom 2024

作者： Huang, Weiguo Ding, Yong Li, Jun Wang, Yujue Liang, Hai Yang, Changsong Guilin University of Electronic Technology Guangxi Key Laboratory of Cryptography and Information Security School of Computer Science and Information Security Guilin China Hkct Institute for Higher Education Lion Rock Labs of Cyberspace Security Institute of Cyberspace Technology 999077 China Guilin University of Electronic Technology Guangxi Engineering Research Center of Industrial Internet Security and Blockchain Guilin China China Industrial Control Systems Cyber Emergency Response Team Beijing China Hangzhou Innovation Institute of Beihang University Hangzhou China

Crowdsourcing testing leverages extensive user participation to identify and fix potential issues in industrial control software, ensuring its security and reliability. Decentralized applications (DApps) can utilize blockchain and smart contract technologies to ensure the decentralization, transparency, and tamper-resistance of industrial control software testing. However, testing efforts conducted via DApps are typically confined to a single blockchain. Coordinating and scheduling DApps deployed on different blockchains has become a critical challenge. Moreover, DApps need to verify the credibility of crowdsourcing participants to ensure the reliability of testing. In order to address these issues, this paper proposes SAMOC, an atomic cross-chain system based on the Atomic-Oracle chain, which aims to realize atomic cross-chain invocations for DApps. SAMOC can realize atomic cross-chain invokes, and the intelligent community deployed on the Atomic-Oracle chain calculates the reputation of users' accounts across multiple chains through Oracle. This paper also designs a corresponding incentive mechanism to ensure that the auditing of cross-chain invokes is honest and trustworthy. Experimental results show that the cross-chain invokes of the proposed SAMOC scheme have lower latency compared to AtomCI. © 2024 IEEE.

关键词： Software testing

WING: Wheel-Inertial Neural Odometry with Ground Manifold Constraints

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Jiang, Chenxing Zhang, Kunyi Yang, Sheng Shen, Shaojie Xu, Chao Gao, Fei The State Key Laboratory of Fluid Power and Mechatronic Systems School of Mechanical Engineering Zhejiang University Hangzhou China State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China Department of Electronic and Computer Engineering Hong Kong University of Science and Technology Hong Kong Alibaba DAMO Academy Hangzhou China Huzhou Institute Zhejiang University Huzhou China

In this paper, we propose an interoceptive-only odometry system for ground robots with neural network processing and soft constraints based on the assumption of a globally continuous ground manifold. Exteroceptive sensors such as cameras, GPS and LiDAR may encounter difficulties in scenarios with poor illumination, indoor environments, dusty areas and straight tunnels. Therefore, improving the pose estimation accuracy only using interoceptive sensors is important to enhance the reliability of navigation system even in degrading scenarios mentioned above. However, interoceptive sensors like IMU and wheel encoders suffer from large drift due to noisy measurements. To overcome these challenges, the proposed system trains deep neural networks to correct the measurements from IMU and wheel encoders, while considering their uncertainty. Moreover, because ground robots can only travel on the ground, we model the ground surface as a globally continuous manifold using a dual cubic B-spline manifold to further improve the estimation accuracy by this soft constraint. A novel space-based sliding-window filtering framework is proposed to fully exploit the C2 continuity of ground manifold soft constraints and fuse all the information from raw measurements and neural networks in a yaw-independent attitude convention. Extensive experiments demonstrate that our proposed approach can outperform state-of-the-art learning-based interoceptive-only odometry methods. © 2024, CC BY.

关键词： Navigation systems

Detection-Triggered Recursive Impact Mitigation against Secondary False Data Injection Attacks in Microgrids

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Liu, Mengxiang Zhang, Xin Zhang, Rui Zhou, Zhuoran Zhang, Zhenyong Deng, Ruilong The Department of Automatic Control and Systems Engineering University of Sheffield Sheffield United Kingdom The National Grid ESO Wokingham United Kingdom The State Key Laboratory of Industrial Control Technology The College of Control Science and Engineering Zhejiang University Hangzhou China

The cybersecurity of microgrid has received widespread attentions due to the frequently reported attack accidents against distributed energy resource (DER) manufactures. Numerous impact mitigation schemes have been proposed to reduce or eliminate the impacts of false data injection attacks (FDIAs). Nevertheless, the existing methods either requires at least one neighboring trustworthy agent or may bring in unacceptable cost burdens. This paper aims to propose a detection-triggered recursive impact mitigation scheme that can timely and precisely counter the secondary FDIAs (SFDIAs) against the communication links among DERs. Once triggering attack alarms, the power line current readings will be utilised to observe the voltage bias injections through the physical interconnections among DERs, based on which the current bias injections can be recursively reconstructed from the residuals generated by unknown input observers (UIOs). The attack impacts are eliminated by subtracting the reconstructed bias from the incoming compromised data. The proposed mitigation method can work even in the worst case where all communication links are under SFDIAs and only require extra current sensors. The bias reconstruction performance under initial errors and system noises is theoretically analysed and the reconstruction error is proved to be bounded regardless of the electrical parameters. To avoid deploying current sensors on all power lines, a cost-effective deployment strategy is presented to secure a spanning tree set of communication links that can guarantee the secondary control performance. Extensive validation studies are conducted in MATLAB/SIMULINK and hardware-in-the-loop (HIL) testbeds to validate the proposed method’s effectiveness against single/multiple and continuous/discontinuous SFDIAs. Copyright © 2024, The Authors. All rights reserved.

关键词： Cost effectiveness

GOOD: General Optimization-based Fusion for 3D Object Detection via LiDAR-Camera Object Candidates

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Shen, Bingqi Dai, Shuwei Chen, Yuyin Xiong, Rong Wang, Yue Jiao, Yanmei The State Key Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou China Hangzhou Iplus Technology Co Ltd Hangzhou China The School of Information Science and Engineering Hangzhou Normal University Hangzhou311121 China

3D object detection serves as the core basis of the perception tasks in autonomous driving. Recent years have seen the rapid progress of multi-modal fusion strategies for more robust and accurate 3D object detection. However, current researches for robust fusion are all learning-based frameworks, which demand a large amount of training data and are inconvenient to implement in new scenes. In this paper, we propose GOOD, a general optimization-based fusion framework that can achieve satisfying detection without training additional models and is available for any combinations of 2D and 3D detectors to improve the accuracy and robustness of 3D detection. First we apply the mutual-sided nearest-neighbor probability model to achieve the 3D-2D data association. Then we design an optimization pipeline that can optimize different kinds of instances separately based on the matching result. Apart from this, the 3D MOT method is also introduced to enhance the performance aided by previous frames. To the best of our knowledge, this is the first optimization-based late fusion framework for multi-modal 3D object detection which can be served as a baseline for subsequent research. Experiments on both nuScenes and KITTI datasets are carried out and the results show that GOOD outperforms by 9.1% on mAP score compared with PointPillars and achieves competitive results with the learning-based late fusion CLOCs. Copyright © 2023, The Authors. All rights reserved.

关键词： Object detection