检索结果-内蒙古大学图书馆

Efficient 3D Deep LiDAR Odometry

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wang, Guangming Wu, Xinrui Jiang, Shuyang Liu, Zhe Wang, Hesheng Department of Automation Key Laboratory of System Control and Information Processing of Ministry of Education Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education Shanghai Engineering Research Center of Intelligent Control and Management Shanghai Jiao Tong University Shanghai200240 China Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai200240 China The Key Laboratory of Artificial Intelligence of Ministry of Education Shanghai Jiao Tong University Shanghai200240 China

An efficient 3D point cloud learning architecture, named EfficientLO-Net, for LiDAR odometry is first proposed in this paper. In this architecture, the projection-aware representation of the 3D point cloud is proposed to organize the raw 3D point cloud into an ordered data form to achieve efficiency. The Pyramid, Warping, and Cost volume (PWC) structure for the LiDAR odometry task is built to estimate and refine the pose in a coarse-to-fine approach. A projection-aware attentive cost volume is built to directly associate two discrete point clouds and obtain embedding motion patterns. Then, a trainable embedding mask is proposed to weigh the local motion patterns to regress the overall pose and filter outlier points. The trainable pose warp-refinement module is iteratively used with embedding mask optimized hierarchically to make the pose estimation more robust for outliers. The entire architecture is holistically optimized end-to-end to achieve adaptive learning of cost volume and mask, and all operations involving point cloud sampling and grouping are accelerated by projection-aware 3D feature learning methods. The superior performance and effectiveness of our LiDAR odometry architecture are demonstrated on KITTI, M2DGR, and Argoverse datasets. Our method outperforms all recent learning-based methods and even the geometry-based approach, LOAM with mapping optimization, on most sequences of KITTI odometry dataset. We open sourced our codes at: https://***/IRMVLab/EfficientLO-Net. Copyright © 2021, The Authors. All rights reserved.

关键词： Iterative methods

Self-supervised discovering of interpretable features for reinforcement learning

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Shi, Wenjie Huang, Gao Song, Shiji Wang, Zhuoyuan Lin, Tingyu Wu, Cheng Department of Automation Tsinghua University Beijing100084 China State Key Laboratory of Intelligent Manufacturing System Technology Beijing Institute of Electronic System Engineering Beijing100039 China

Deep reinforcement learning (RL) has recently led to many breakthroughs on a range of complex control tasks. However, the agent's decision-making process is generally not transparent. The lack of interpretability hinders the applicability of RL in safety-critical scenarios. While several methods have attempted to interpret vision-based RL, most come without detailed explanation for the agent's behavior. In this paper, we propose a self-supervised interpretable framework, which can discover interpretable features to enable easy understanding of RL agents even for non-experts. Specifically, a self-supervised interpretable network (SSINet) is employed to produce fine-grained attention masks for highlighting task-relevant information, which constitutes most evidence for the agent's decisions. We verify and evaluate our method on several Atari 2600 games as well as Duckietown, which is a challenging self-driving car simulator environment. The results show that our method renders empirical evidences about how the agent makes decisions and why the agent performs well or badly, especially when transferred to novel scenes. Overall, our method provides valuable insight into the internal decision-making process of vision-based RL. In addition, our method does not use any external labelled data, and thus demonstrates the possibility to learn high-quality mask through a self-supervised manner, which may shed light on new paradigms for label-free vision learning such as self-supervised segmentation and detection. Copyright © 2020, The Authors. All rights reserved.

关键词： Deep learning

3D Hierarchical Refinement and Augmentation for Unsupervised Learning of Depth and Pose from Monocular Video

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wang, Guangming Zhong, Jiquan Zhao, Shijie Wu, Wenhua Liu, Zhe Wang, Hesheng Department of Automation Institute of Medical Robotics Key Laboratory of System Control and Information Processing Ministry of Education Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education Shanghai Engineering Research Center of Intelligent Control and Management Shanghai Jiao Tong University Shanghai200240 China Department of Engineering Mechanics Shanghai Jiao Tong University Shanghai200240 China Department of Computer Science and Technology University of Cambridge CambridgeCB2 1TN United Kingdom

Depth and ego-motion estimations are essential for the localization and navigation of autonomous robots and autonomous driving. Recent studies make it possible to learn the per-pixel depth and ego-motion from the unlabeled monocular video. A novel unsupervised training framework is proposed with 3D hierarchical refinement and augmentation using explicit 3D geometry. In this framework, the depth and pose estimations are hierarchically and mutually coupled to refine the estimated pose layer by layer. The intermediate view image is proposed and synthesized by warping the pixels in an image with the estimated depth and coarse pose. Then, the residual pose transformation can be estimated from the new view image and the image of the adjacent frame to refine the coarse pose. The iterative refinement is implemented in a differentiable manner in this paper, making the whole framework optimized uniformly. Meanwhile, a new image augmentation method is proposed for the pose estimation by synthesizing a new view image, which creatively augments the pose in 3D space but gets a new augmented 2D image. The experiments on KITTI demonstrate that our depth estimation achieves state-of-the-art performance and even surpasses recent approaches that utilize other auxiliary tasks. Our visual odometry outperforms all recent unsupervised monocular learning-based methods and achieves competitive performance to the geometry-based method, ORB-SLAM2 with back-end optimization. Copyright © 2021, The Authors. All rights reserved.

关键词： Robots

Semi-Global Consensus Problems of Discrete-Time Multi-Agent systems in the Presence of Input Constraints

学校读者我要写书评

暂无评论

Journal of Shanghai Jiaotong university(Science) 2020年第3期25卷 288-298页

作者： WANG Jian SHI Liangren Department of Automation Shanghai Jiao Tong UniversityShanghai 200240China Key Laboratory of System Control and Information Processing of Ministry of Education Shanghai Jiao Tong UniversityShanghai 200240China Shanghai Key Laboratory of Navigation and Location-Based Services Shanghai Jiao Tong UniversityShanghai 200240China Academy of Information Technology and Electrical Engineering Shanghai Jiao Tong UniversityShanghai 200240China

This paper studies the consensus problems of the discrete-time multi-agent systems (MASs) in the presence of input saturation constraints over directed communication *** kinds of consensus problems,of the leaderless consensus problem with no leader agent and the containment control problem with multiple leader agents,are investigated in this *** gain feedback consensus algorithms based on the improved discretetime parametric algebraic Riccati equation (ARE) are proposed to solve the consensus *** the MAS without any leader,the trajectories of all agents converge together when the communication networks contain a directed spanning *** the MAS with multiple leaders,the trajectories of all follower agents converge to the convex hull spanned by the leader agents when there is at least one leader agent which is trackable for each follower *** examples are provided to verify the theoretical results.

关键词： leaderless consensus containment control discrete-time low gain input saturation

Topology inference for multi-agent cooperation under unmeasurable latent input

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Jiao, Qing Li, Yushan He, Jianping Shi, Ling Department of Automation Shanghai Jiao Tong University Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai China Department of Electronic and Computer Engineering Hong Kong University of Science and Technology Hong Kong

Topology inference is a crucial problem for cooperative control in multi-agent systems. Different from most prior works, this paper is dedicated to inferring the directed network topology from the observations that consist of a single, noisy and finite time-series system trajectory, where the cooperation dynamics is stimulated with the initial network state and the unmeasurable latent input. The unmeasurable latent input refers to intrinsic system signal and extrinsic environment interference. Considering the time-invariant/varying nature of the input, we propose two-layer optimization-based and iterative estimation-based topology inference algorithms (TO-TIA and IE-TIA), respectively. TO-TIA allows us to capture the separability of global agent state and eliminates the unknown influence of time-invariant input on system dynamics. IE-TIA further exploits the identifiability and estimability of more general time-varying input and provides an asymptotic solution with desired convergence properties, with higher computation cost compared with TO-TIA. Our novel algorithms relax the dependence of observation scale and leverage the empirical risk reformulation to improve the inference accuracy in terms of the topology structure and edge weight. Comprehensive theoretical analysis and simulations for various topologies are provided to illustrate the inference feasibility and the performance of the proposed algorithms. Copyright © 2020, The Authors. All rights reserved.

关键词： Multi agent systems

Reliability Analysis of Assembly Processes Performed by Human-Robot Interaction 3

学校读者我要写书评

暂无评论

Reliability Analysis of Assembly Processes Performed by Huma...

3rd International Symposium on Multidisciplinary Studies and Innovative Technologies, ISMSIT 2019

作者： Coban, Mustafa Kaymakci, Ozgur Turay Gelen, Gokhan Bursa Technical University Department of Mechatronics Engineering Bursa Turkey Yildiz Technical University Faculty of Electric Electronic Control and Automation Engineering Istanbul Turkey

ISBN: (纸本)9781728137896

Together with innovations in robot technologies and the emergence of the fourth industrial revolution called Industry 4.0, collaborative robot applications have improved. Collaborative robots are the robots that provide the safe handling of many processes such as assembly, pick and place, palletizing in factories. The reliability analysis of the system has great importance in order to determine the failure conditions and failure possibilities that will occur in the systems enabling these operations. In this study, reliability analysis of assembly processes performed by human-robot interaction is presented. The reliability block diagram of the system is created and the reliability value of the system over time is determined. Then, Markov model indicating the fault conditions in the system is created and the probability of occurrence of these fault conditions is determined. The results show that if the system operates continuously for five years, the reliability will drop below 10% and the reasons that decrease the reliability are discussed in the last section. © 2019 IEEE.

关键词： Industrial robots

Novel Diffusion Models for Multimodal 3D Hand Trajectory Prediction

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Ma, Junyi Bao, Wentao Xu, Jingyi Sun, Guanzhong Chen, Xieyuanli Wang, Hesheng IRMV Lab The Department of Automation Shanghai Jiao Tong University China Meta Reality Labs United States Department of Electronic Engineering Shanghai Jiao Tong University China School of Information and Control Engineering China University of Mining and Technology China College of Intelligence Science and Technology National University of Defense Technology China

Predicting hand motion is critical for understanding human intentions and bridging the action space between human movements and robot manipulations. Existing hand trajectory prediction (HTP) methods forecast the future hand waypoints in 3D space conditioned on past egocentric observations. However, such models are only designed to accommodate 2D egocentric video inputs. There is a lack of awareness of multimodal environmental information from both 2D and 3D observations, hindering the further improvement of 3D HTP performance. In addition, these models overlook the synergy between hand movements and headset camera egomotion, either predicting hand trajectories in isolation or encoding egomotion only from past frames. To address these limitations, we propose novel diffusion models (MMTwin) for multimodal 3D hand trajectory prediction. MMTwin is designed to absorb multimodal information as input encompassing 2D RGB images, 3D point clouds, past hand waypoints, and text prompt. Besides, two latent diffusion models, the egomotion diffusion and the HTP diffusion as twins, are integrated into MMTwin to predict camera egomotion and future hand trajectories concurrently. We propose a novel hybrid Mamba-Transformer module as the denoising model of the HTP diffusion to better fuse multimodal features. The experimental results on three publicly available datasets and our self-recorded data demonstrate that our proposed MMTwin can predict plausible future 3D hand trajectories compared to the state-of-the-art baselines, and generalizes well to unseen environments. The code and pretrained models will be released at https://***/IRMVLab/MMTwin. © 2025, CC BY-NC-SA.

关键词：

A Fine Grained Forecasting Method for Minute-Level Urban Water Demand

学校读者我要写书评

暂无评论

A Fine Grained Forecasting Method for Minute-Level Urban Wat...

第40届中国控制会议

作者： Shunyu Wu Pingwei Zhao Miaoshun Bai Jingcheng Wang Yang Lan Department of Automation Shanghai Jiao Tong University Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai Chengtou Water Group Co. LTD. Shanghai Municipal Engineering Design Institute Group Co. Ltd.

Forecasting of one day’s Minute-level urban water demand is essential for urban water *** forecasting in one day requires super long steps which are around *** multi-step forecasting method is almost impossiable to achieve the prediction of ultra-long step *** this paper,we propose a multi-step forecasting method named Fine Grained Forecasting（FGF）.Specifically,this method firstly forecasts cumulative flow of urban water demand which is daily water *** fine grained division,searching and prediction are applied for minute-level urban water *** principle that FGF is effective has been proved to be valid by careful *** a LSTM-Self-Attention deep neural network is proposed to realize fast and accurate forecasting of daily urban water *** results of daily and minute-level urban water demand have been proved to achieve high accuracy through *** shows the effectiveness of fine grained forecasting.

关键词： Multi-step forecasting Fine grained forecasting LSTM Self-attention

Distributed urban freeway traffic optimization considering congestion propagation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Gao, Fengkun Yang, Bo Chen, Cailian Guan, Xinping Zhang, Yang Department of Automation Shanghai Jiao Tong University Shanghai200240 China Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai200240 China Shanghai Engineering Research Center of Industrial Intelligent Control and Management Shanghai200240 China Shanghai Municipal Urban-Rural Construction and Transportation Development Institute China

Traffic optimization strategies are imperative for improving the performance of transportation networks. Most traffic optimization strategies only depend on traffic states of congested road segments, where congestion propagation is neglected. Therefore, we propose a distributed traffic optimization strategy for urban freeways considering the potential congested road segments caused by congestion propagation, called potential-homogeneous-area (PHA). Utilizing the historical traffic density data, we firstly quantify the effect of congestion propagation and identify PHA by applying the proposed spatio-temporal lambda-connectedness method. Meanwhile, dynamic capacity constraints of PHA are determined and are integrated with cell transmission model (CTM) in a centralized traffic optimization problem. To reduce computational complexity and improve scalability, we then propose a doubleconsensus- based alternating direction method of multipliers algorithm (DC-ADMM) to handle the neighbor coupling constraints and global coupling constraints for solving the problem in a fully distributed way. We prove that the proposed DC-ADMM algorithm converges to the optimal solution in the condition of convex objective function. Finally, simulations based on real traffic density data, collected in Inner Ring Road, Shanghai, China, reveal the effectiveness of our proposed strategy. © 2021, CC BY.

关键词： Roads and streets