检索结果-内蒙古大学图书馆

Revisiting 2D convolutional neural networks for graph-based applications

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Lyu, Yecheng Huang, Xinming Zhang, Ziming The Department of Electrical and Computer Engineering Worcester Polytechnic Institute WorcesterMA01609 United States

Graph convolutional networks (GCNs) are widely used in graph-based applications such as graph classification and segmentation. However, current GCNs have limitations on implementation such as network architectures due to their irregular inputs. In contrast, convolutional neural networks (CNNs) are capable of extracting rich features from large-scale input data, but they do not support general graph inputs. To bridge the gap between GCNs and CNNs, in this paper we study the problem of how to effectively and efficiently map general graphs to 2D grids that CNNs can be directly applied to, while preserving graph topology as much as possible. We therefore propose two novel graph-to-grid mapping schemes, namely, graph-preserving grid layout (GPGL) and its extension Hierarchical GPGL (H-GPGL) for computational efficiency. We formulate the GPGL problem as integer programming and further propose an approximate yet efficient solver based on a penalized Kamada-Kawai method, a well-known optimization algorithm in 2D graph drawing. We propose a novel vertex separation penalty that encourages graph vertices to lay on the grid without any overlap. Along with this image representation, even extra 2D maxpooling layers contribute to the PointNet, a widely applied point-based neural network. We demonstrate the empirical success of GPGL on general graph classification with small graphs and H-GPGL on 3D point cloud segmentation with large graphs, based on 2D CNNs including VGG16, ResNet50 and multi-scale maxout (MSM) CNN. © 2021, CC BY.

关键词： Integer programming

Virtual Regression-Based Myoelectric Hand-Wrist Prosthesis Control and Electrode Site Selection Using No Force Feedback

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Li, Jianan Zhu, Ziling Boyd, William J. Martinez-Luna, Carlos Dai, Chenyun Wang, Haopeng Wang, He Huang, Xinming Farrell, Todd R. Clancy, Edward A. Department of Electrical and Computer Engineering Worcester Polytechnic Institute WorcesterMA01609 United States Liberating Technologies Inc. HollistonMA01746 United States Department of Electrical Engineering Fudan University Shanghai200433 China

Most transradial prosthesis users with conventional 'Sequential' myoelectric control have two electrode sites which control one degree of freedom (DoF) at a time. Rapid EMG co-activation toggles control between DoFs (e.g., hand and wrist), providing limited function. We implemented a regressionbased EMG control method which achieved simultaneous and proportional control of two DoFs in a virtual task. We automated electrode site selection using a short-duration (90 s) calibration period, without force feedback. Backward stepwise selection located the best electrodes for either six or 12 electrodes (selected from a pool of 16). We additionally studied two, 2-DoF controllers: 'Intuitive' control (hand open-close and wrist pronation-supination controlled virtual target size and rotation, respectively) and 'Mapping' control (wrist flexion-extension and ulnar-radial deviation controlled virtual target left-right and up-down movement, respectively). In practice, a Mapping controller would be mapped to control prosthesis hand open-close and wrist pronation-supination. Eleven able-bodied subjects and 4 limb-absent subjects completed virtual target matching tasks (fixed target moves to a new location after being 'matched,' and subject immediately pursues) and fixed (static) target tasks. For all subjects, both 2-DoF controllers with 6 optimally-sited electrodes had statistically better target matching performance than Sequential control in number of matches (average of 4–7 vs. 2 matches, p © 2022, The Authors. All rights reserved.

关键词： Electrodes

Verification and synthesis of control barrier functions

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Clark, Andrew Department of Electrical and Computer Engineering Worcester Polytechnic Institute 100 Institute Road WorcesterMA01609 United States

Control systems often must satisfy strict safety requirements over an extended operating lifetime. Control Barrier Functions (CBFs) are a promising recent approach to constructing simple and safe control policies. This paper proposes a framework for verifying that a CBF guarantees safety for all time and synthesizing CBFs with verifiable safety in polynomial control systems. Our approach is to show that safety of CBFs is equivalent to the non-existence of solutions to a family of polynomial equations, and then prove that this nonexistence is equivalent to a pair of sum-of-squares constraints via the Positivstellensatz of algebraic geometry. We develop this Positivstellensatz to verify CBFs, as well as generalization to high-degree systems and multiple CBF constraints. We then propose a set of heuristics for CBF synthesis, including a general alternating-descent heuristic, a specialized approach for compact safe regions, and an approach for convex unsafe regions. Our approach is illustrated on two numerical examples. © 2021, CC BY.

关键词： Polynomials

Safe Control for Nonlinear Systems under Faults and Attacks via Control Barrier Functions

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Zhang, Hongchao Li, Zhouchi Clark, Andrew The Electrical and Systems Engineering Department McKelvey School of Engineering Washington University in St. Louis St. LouisMO63130 United States The Department of Electrical and Computer Engineering Worcester Polytechnic Institute WorcesterMA01609 United States

Safety is one of the most important properties of control systems. Sensor faults and attacks and actuator failures may cause errors in the sensor measurements and system dynamics, which leads to erroneous control inputs and hence safety violations. In this paper, we improve the robustness against sensor faults and actuator failures by proposing a class of Fault-Tolerant Control Barrier Functions (FT-CBFs) for nonlinear systems. Our approach maintains a set of state estimators according to fault patterns and incorporates CBF-based linear constraints for each state estimator. We then propose a framework for joint safety and stability by integrating FT-CBFs with Control Lyapunov Functions. With a similar philosophy of utilizing redundancy, we proposed High order CBF-based approach to ensure safety when actuator failures occur. We propose a sum-of-squares (SOS) based approach to verify the feasibility of FT-CBFs for both sensor faults and actuator failures. We evaluate our approach via two case studies, namely, a wheeled mobile robot (WMR) system in the presence of a sensor attack and a Boeing 747 lateral control system under actuator failures. Copyright © 2022, The Authors. All rights reserved.

关键词： Nonlinear systems

A divide-and-merge point cloud clustering algorithm for LiDAR panoptic segmentation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhao, Yiming Zhang, Xiao Huang, Xinming Department of Electrical and Computer Engineering Worcester Polytechnic Institute Massachusetts01609 United States

Clustering objects from the LiDAR point cloud is an important research problem with many applications such as autonomous driving. To meet the real-time requirement, existing research proposed to apply the connected-componentlabeling (CCL) technique on LiDAR spherical range image with a heuristic condition to check if two neighbor points are connected. However, LiDAR range image is different from a binary image which has a deterministic condition to tell if two pixels belong to the same component. The heuristic condition used on the LiDAR range image only works empirically, which suggests the LiDAR clustering algorithm should be robust to potential failures of the empirical heuristic condition. To overcome this challenge, this paper proposes a divide-and-merge LiDAR clustering algorithm. This algorithm firstly conducts clustering in each evenly divided local region, then merges the local clustered small components by voting on edge point pairs. Assuming there are N LiDAR points of objects in total with m divided local regions, the time complexity of the proposed algorithm is O(N) + O(m2). A smaller m means the voting will involve more neighbor points, but the time complexity will become larger. So the m controls the trade-off between the time complexity and the clustering accuracy. A proper m helps the proposed algorithm work in real-time as well as maintain good performance. We evaluate the divide-andmerge clustering algorithm on the SemanticKITTI panoptic segmentation benchmark by cascading it with a state-of-the-art semantic segmentation model. The final performance evaluated through the leaderboard achieves the best among all published methods. The proposed algorithm is implemented with C++ and wrapped as a python function. It can be easily used with the modern deep learning framework in python. We will release the code under the following link. Copyright © 2021, The Authors. All rights reserved.

关键词： Economic and social effects

FIDNet: LiDAR point cloud semantic segmentation with fully interpolation decoding

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhao, Yiming Bai, Lin Huang, Xinming Department of Electrical and Computer Engineering Worcester Polytechnic Institute MA01609 United States

Projecting the point cloud on the 2D spherical range image transforms the LiDAR semantic segmentation to a 2D segmentation task on the range image. However, the LiDAR range image is still naturally different from the regular 2D RGB image;for example, each position on the range image encodes the unique geometry information. In this paper, we propose a new projection-based LiDAR semantic segmentation pipeline that consists of a novel network structure and an efficient post-processing step. In our network structure, we design a FID (fully interpolation decoding) module that directly upsamples the multi-resolution feature maps using bilinear interpolation. Inspired by the 3D distance interpolation used in PointNet++, we argue this FID module is a 2D version distance interpolation on (θ, φ) space. As a parameter-free decoding module, the FID largely reduces the model complexity by maintaining good performance. Besides the network structure, we empirically find that our model predictions have clear boundaries between different semantic classes. This makes us rethink whether the widely used K-nearest-neighbor post-processing is still necessary for our pipeline. Then, we realize the many-to-one mapping causes the blurring effect that some points are mapped into the same pixel and share the same label. Therefore, we propose to process those occluded points by assigning the nearest predicted label to them. This NLA (nearest label assignment) post-processing step shows a better performance than KNN with faster inference speed in the ablation study. On SemanticKITTI dataset, our pipeline achieves the best performance among all projection-based methods with 64 × 2048 resolution and all point-wise solutions. With a ResNet-34 as the backbone, both the training and testing of our model can be finished on a single RTX 2080 Ti with 11G memory. The code is released here. Copyright © 2021, The Authors. All rights reserved.

关键词： Pipelines

A surface geometry model for LiDAR depth completion

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zhao, Yiming Bai, Lin Zhang, Ziming Huang, Xinming Department of Electrical and Computer Engineering Worcester Polytechnic Institute MA01609 United States

— LiDAR depth completion is a task that predicts depth values for every pixel on the corresponding camera frame although only sparse LiDAR points are available. Most of the existing state-of-the-art solutions are based on deep neural networks, which need a large amount of data and heavy computations for training the models. In this letter, a novel non-learning depth completion method is proposed by exploiting the local surface geometry that is enhanced by an outlier removal algorithm. The proposed surface geometry model is inspired by the observation that most pixels with unknown depth have a nearby LiDAR point. Therefore, it is assumed those pixels share the same surface with the nearest LiDAR point, and their respective depth can be estimated as the nearest LiDAR depth value plus a residual error. The residual error is calculated by using a derived equation with several physical parameters as input, including the known camera intrinsic parameters, estimated normal vector, and offset distance on the image plane. The proposed method is further enhanced by an outlier removal algorithm that is designed to remove incorrectly mapped LiDAR points from occluded regions. On KITTI dataset, the proposed solution achieves the best error performance among all existing non-learning methods and is comparable to the best self-supervised learning method and some supervised learning methods. Moreover, since removing outlier points coming from occluded regions is a commonly existing problem in LiDAR-Camera systems, the proposed outlier removal algorithm is a general preprocessing step that is applicable to many robotic systems with both camera and LiDAR sensors. 1 Copyright © 2021, The Authors. All rights reserved.

关键词： Pixels

B-COLORINGS OF PLANAR AND OUTERPLANAR GRAPHS

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Martin, Ryan R. Ruszinkó, Miklós Sárközy, Gábor N. Department of Mathematics Iowa State University AmesIA United States HUN-REN Alfréd Rényi Institute of Mathematics Budapest Hungary Department of Sciences and Engineering Sorbonne University Abu Dhabi Abu Dhabi United Arab Emirates Faculty of Information Technology and Bionics Pázmány Péter Catholic University Hungary Computer Science Department Worcester Polytechnic Institute WorcesterMA United States

A coloring of the edges of a graph G in which every K1,2 is totally multicolored is known as a proper coloring and a coloring of the edges of G in which every K1,2 and every K2,2 is totally multicolored is called a B-coloring. In this paper, we establish that a planar graph with maximum degree ∆ can be B-colored with max{2∆, 32} colors. This is best-possible for large ∆ because K2,∆ requires 2∆ colors. In addition, there is an example with ∆ = 4 that requires 12 colors. We also establish that an outerplanar graph with maximum degree ∆ can be B-colored with max{∆, 6} colors. This is almost best-possible because ∆ colors are necessary and there is an example with ∆ = 4 that requires 5 *** Codes 05C15, 05C10, 05C35 © 2024, CC BY.

关键词：

An Analytical Framework for Control Synthesis of Cyber-Physical Systems with Safety Guarantee

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Niu, Luyao Al Maruf, Abdullah Clark, Andrew Mertoguno, J. Sukarno Poovendran, Radha The Department of Electrical and Computer Engineering Worcester Polytechnic Institute WorcesterMA01609 United States The Network Security Lab Department of Electrical and Computer Engineering University of Washington SeattleWA98195-2500 United States Information and Cyber Sciences Georgia Tech Research Institute AtlantaGA30332 United States

Cyber-physical systems (CPS) are required to operate safely under fault and malicious attacks. The simplex architecture and the recently proposed cyber resilient architectures, e.g., Byzantine fault tolerant++ (BFT++), provide safety for CPS under faults and malicious cyber attacks, respectively. However, these existing architectures make use of different timing parameters and implementations to provide safety, and are seemingly unrelated. In this paper, we propose an analytical framework to represent the simplex, BFT++ and other practical cyber resilient architectures (CRAs). We construct a hybrid system that models CPS adopting any of these architectures. We derive sufficient conditions via our proposed framework under which a control policy is guaranteed to be safe. We present an algorithm to synthesize the control policy. We validate the proposed framework using a case study on lateral control of a Boeing 747, and demonstrate that our proposed approach ensures safety of the system. © 2022, CC BY.

关键词： Embedded systems