检索结果-内蒙古大学图书馆

Addressing Sample Efficiency and Model-bias in Model-based Reinforcement Learning

学校读者我要写书评

暂无评论

Addressing Sample Efficiency and Model-bias in Model-based R...

International Conference on Machine Learning and Applications (ICMLA)

作者： Akhil S Anand Jens Erik Kveen Fares Abu-Dakka Esten Ingar Grøtli Jan Tommy Gravdahl Dept. of Engineering Cybernetics Norwegian University of Science and Technology (NTNU) Trondheim Norway Department of Electrical Engineering and Automation (EEA) Intelligent Robotics Group Aalto University Aalto Finland Dept. of Mathematics and Cybernetics SINTEF Digital Trondheim Norway

Model-based reinforcement learning promises to be an effective way to bring reinforcement learning to real-world robotic systems by offering a sample efficient learning approach compared to model-free reinforcement learning. However, model-based reinforcement learning approaches at present struggle to match the performance of model-free ones. This work attempts to fill this gap by improving the performance of model-based reinforcement learning while further improving its sample efficiency. To improve the sample efficiency, an exploration strategy is formulated which maximizes the information gain. The asymptotic performance is improved by compensating for the model-bias using a model-free critic. We have evaluated our proposed approach on four reinforcement learning benchmarking tasks in the openAI gym framework.

关键词： Reinforcement learning Benchmark testing Data models Task analysis Robots

A Unified Formulation of Geometry-aware Discrete Dynamic Movement Primitives

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Abu-Dakka, Fares J. Saveriano, Matteo Kyrki, Ville Electronic and Computer Science Department Faculty of Engineering Mondragon Unibertsitatea Arrasate20500 Spain Automatic Control Lab Department of Industrial Engineering University of Trento Trento Italy Intelligent Robotics Group Department of Electrical Engineering and Automation Aalto University Aalto Finland

Learning from demonstration (LfD) is considered as an efficient way to transfer skills from humans to robots. Traditionally, LfD has been used to transfer Cartesian and joint positions and forces from human demonstrations. The traditional approach works well for some robotic tasks, but for many tasks of interest, it is necessary to learn skills such as orientation, impedance, and/or manipulability that have specific geometric characteristics. An effective encoding of such skills can be only achieved if the underlying geometric structure of the skill manifold is considered and the constrains arising from this structure are fulfilled during both learning and execution. However, typical learned skill models such as dynamic movement primitives (DMPs) are limited to Euclidean data and fail in correctly embedding quantities with geometric constraints. In this paper, we propose a novel and mathematically principled framework that uses concepts from Riemannian geometry to allow DMPs to properly embed geometric constrains. The resulting DMP formulation can deal with data sampled from any Riemannian manifold including, but not limited to, unit quaternions and symmetric and positive definite matrices. The proposed approach has been extensively evaluated both on simulated data and real robot experiments. The performed evaluation demonstrates that beneficial properties of DMPs, such as convergence to a given goal and the possibility to change the goal during operation, apply also to the proposed formulation. © 2022, CC BY-NC-ND.

关键词： Demonstrations

Roof Supports Measurement in Coal Mining by 3D Registration and Model Fitting

学校读者我要写书评

暂无评论

Roof Supports Measurement in Coal Mining by 3D Registration ...

2021 China automation Congress, CAC 2021

作者： Dong, Yan Zhang, Tao Wang, Yongqiang Zhang, Xingfu Zhu, Jun Yang, Xin Han, Bin State Key Lab of Digital Manufacturing Equipment and Technology Huazhong University of Science and Technology Wuhan430074 China Department of Automation Tsinghua University Beijing100084 China Zhengzhou Coal Mining Machinery Group Co. Ltd. Zhengzhou450016 China Guangdong Intelligent Robotics Institute Dongguan523808 China

ISBN: (纸本)9781665426473

The roof support, a type of special equipment for coal mining, protect miners and other facilities in fully mechanized working face under the ground, and move with shearers when extracting coal. The states of each support, such as the position, height and tilt angle are important for its motion control. Unfortunately, it is difficult to measure these parameters directly in the working face. The novelty of this paper is, we use two LiDARs carried by a rail robot to scan the whole working face for the point cloud and adopt 3D registration and model fitting methods to calculate the key parameters of supports. The experiments show that our method can successfully tackle this problem. Since there are few works that used LiDARs for support measurement, we believe our effort would contribute to the intelligent coal mining. © 2021 IEEE

关键词： Roofs

DINF: Dynamic Instance Noise Filter for Occluded Pedestrian Detection

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Li, Xiang He, Miao Luo, Haibo Xiao, Jiajie Key Laboratory of Opto-Electronic Information Processing Chinese Academy of Sciences Liaoning Shenyang110016 China Shenyang Institute of Automation Chinese Academy of Sciences Liaoning Shenyang110016 China Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Liaoning Shenyang110169 China University of Chinese Academy of Sciences Beijing100049 China *** Group Shanghai200335 China

Occlusion issue is the biggest challenge in pedestrian detection. RCNN-based detectors extract instance features by cropping rectangle regions of interest in the feature maps. However, the visible pixels of the occluded objects are limited, making the rectangle instance feature mixed with a lot of instance-irrelevant noise information. Besides, by counting the number of instances with different degrees of overlap of CrowdHuman dataset, we find that the number of severely overlapping objects and slightly overlapping objects are unbalanced, which may exacerbate the challenges posed by occlusion *** to the noise issue, from the perspective of denoising, an iterable dynamic instance noise filter (DINF) is proposed for the RCNN-based pedestrian detectors to improve the signal-noise ratio of the instance feature. Simulating the wavelet denoising process, we use the instance feature vector to generate dynamic convolutional kernels to transform the RoIs features to a domain in which the near-zero values represent the noise information. Then, soft thresholding with channel-wise adaptive thresholds is applied to convert the near-zero values to zero to filter out noise information. For the imbalance issue, we propose an IoU-Focal factor (IFF) to modulate the contributions of the well-regressed boxes and the bad-regressed boxes to the loss in the training process, paying more attention to the minority severely overlapping objects. Extensive experiments conducted on CrowdHuman and CityPersons demonstrate that our methods can help RCNN-based pedestrian detectors achieve state-of-the-art performance. Copyright © 2023, The Authors. All rights reserved.

关键词： Object detection

Enhancing Robot Teleoperation in Remote automation Production through an Event-Triggered Control Strategy

学校读者我要写书评

暂无评论

Enhancing Robot Teleoperation in Remote Automation Productio...

IEEE International Symposium on Industrial Electronics (ISIE)

作者： Mengke Wang Honghao Lv Ruohan Wang Haiteng Wu Lipeng Chen Yi Chen Haihui Yuan Geng Yang State Key Laboratory of Fluid Power and Mechatronic Systems School of Mechanical Engineering Zhejiang University Hangzhou China Zhejiang Key Laboratory of Intelligent Operation and Maintenance Robot Hangzhou Shenhao Technology Hangzhou China Robotics X Tencent Shenzhen China Intelligent Manufacturing Center DongFang Electric Corporation Academy of Science and Technology Co Chengdu China China South Industries Group Corporation Hangzhou Automation Research Insititute Co. Ltd. China State Key Laboratory of Fluid Power and Mechatronic Systems School of Mechanical Engineering Zhejiang University Zhejiang Key Laboratory of Intelligent Operation and Maintenance Robot Hangzhou China

ISBN: (数字)9798350394085

ISBN: (纸本)9798350394092

With the advancement of intelligent manufacturing, an increasing number of industrial production processes are transforming unmanned operations. This trend is particularly evident in factories that involve potential safety hazards, where the demand for unmanned production is becoming increasingly urgent. Teleoperation, as a core technology for achieving unmanned factories, holds tremendous importance in keeping workers safe from danger. However, traditional teleoperation strategies encounter challenges due to mismatches between master and slave workspaces, making it difficult to achieve both high precision and operation across large workspaces simultaneously. To address this issue, this paper proposes an event-triggered control strategy for a teleoperated robot system. By introducing a foot pedal as the event-triggered switch, the enablement and termination of teleoperation are associated with the activation of the pedal. Through resetting the start position, fine-grained teleoperation is achieved throughout the entire workspace of the robot. To demonstrate the advantages of the proposed control strategy, a mimicked wiring operation task scenario is constructed for experimental verification, and the teleoperation tasks are completed successfully and efficiently.

关键词： Wiring Accuracy Service robots Switches Market research Production facilities Main-secondary

Periodic DMP formulation for quaternion trajectories

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Abu-Dakka, Fares J. Saveriano, Matteo Peternel, Luka Intelligent Robotics Group Dept of Electrical Engineering and Automation Aalto University Finland Dept of Computer Science and Digital Science Center University of Innsbruck Austria Delft Haptics Lab Dept of Cognitive Robotics Delft University of Technology Delft Netherlands

Imitation learning techniques have been used as a way to transfer skills to robots. Among them, dynamic movement primitives (DMPs) have been widely exploited as an effective and an efficient technique to learn and reproduce complex discrete and periodic skills. While DMPs have been properly formulated for learning point-to-point movements for both translation and orientation, periodic ones are missing a formulation to learn the orientation. To address this gap, we propose a novel DMP formulation that enables encoding of periodic orientation trajectories. Within this formulation we develop two approaches: Riemannian metric-based projection approach and unit quaternion based periodic DMP. Both formulations exploit unit quaternions to represent the orientation. However, the first exploits the properties of Riemannian manifolds to work in the tangent space of the unit sphere. The second encodes directly the unit quaternion trajectory while guaranteeing the unitary norm of the generated quaternions. We validated the technical aspects of the proposed methods in simulation. Then we performed experiments on a real robot to execute daily tasks that involve periodic orientation changes (i.e., surface polishing/wiping and liquid mixing by shaking). © 2021, CC BY.

关键词： Trajectories

TSAA: A Two-Stage Anchor Assignment Method towards Anchor Drift in Crowded Object Detection

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Xiang, Li Miao, He Haibo, Luo Huiyuan, Yang Jiajie, Xiao Key Laboratory of Opto-Electronic Information Processing Chinese Academy of Sciences China Shenyang Institute of Automation Chinese Academy of Sciences China Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences China University of Chinese Academy of Sciences China McGill University Montreal Canada *** Group China

Among current anchor-based detectors, a positive anchor box will be intuitively assigned to the object that overlaps it the most. The assigned label to each anchor will directly determine the optimization direction of the corresponding prediction box, including the direction of box regression and category prediction. In our practice of crowded object detection, however, the results show that a positive anchor does not always regress toward the object that overlaps it the most when multiple objects overlap. We name it anchor drift. The anchor drift reflects that the anchor-object matching relation, which is determined by the degree of overlap between anchors and objects, is not always optimal. Conflicts between the fixed matching relation and learned experience in the past training process may cause ambiguous predictions and thus raise the false-positive rate. In this paper, a simple but efficient adaptive two-stage anchor assignment (TSAA) method is proposed. It utilizes the final prediction boxes rather than the fixed anchors to calculate the overlap degree with objects to determine which object to regress for each anchor. The participation of the prediction box makes the anchor-object assignment mechanism adaptive. Extensive experiments are conducted on three classic detectors RetinaNet, Faster-RCNN and YOLOv3 on CrowdHuman and COCO to evaluate the effectiveness of TSAA. The results show that TSAA can significantly improve the detectors’ performance without additional computational costs or network structure changes. Copyright © 2022, The Authors. All rights reserved.

关键词： Anchors

Automatic installation system of soft rubber caps based on multi-rivet aircraft structural parts

学校读者我要写书评

暂无评论

Automatic installation system of soft rubber caps based on m...

Electrical Engineering and automation (IFEEA), International Forum on

作者： Qingshi Zhang Jianwei Mu Bowen Xu Chuanjun Yu Xin Tong Wei Wang Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang China Key Laboratory on Intelligent Detection and Equipment Technology of Liaoning Province Shenyang China AVIC Shenyang Aircraft Industrial (Group) Co. Ltd Shenyang China School of Mechanical Engineering Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9781665464222

Many structural parts are connected and fixed with rivets or screws in aircraft assembly. Due to the requirements of sealing, anticorrosion or fastening, soft rubber caps coated with sealant need to be bonded. The current production mode is mostly manual operation, there are a lot of problems, such as heavy workload, high hollow rate of sealing cap, low production efficiency, not environmental-friendly and so on. In this paper, a multi-rivet automatic identification and rubber cap installation system of micro-curved structural parts based on robotic linear laser scanning is designed by cooperative-robot to carry a linear laser camera and a soft rubber cap pickup fixture. The system can realize the automatic scanning, identification, classification and positioning of rivets or screws, automatically pick up different types of rubber caps according to the classification, and link with the automatic glue injection system to complete the quantitative glue injection of rubber caps, and automatically complete the installation of rubber caps according to the positioning information.

关键词： Sealing materials Visualization Semiconductor lasers Production Fasteners Rubber Aircraft

Geometry-aware Dynamic Movement Primitives

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Abu-Dakka, Fares J. Kyrki, Ville Intelligent Robotics Group Department of Electrical Engineering and Automation [EEA Aalto University Espoo Finland

— In many robot control problems, factors such as stiffness and damping matrices and manipulability ellipsoids are naturally represented as symmetric positive definite (SPD) matrices, which capture the specific geometric characteristics of those factors. Typical learned skill models such as dynamic movement primitives (DMPs) can not, however, be directly employed with quantities expressed as SPD matrices as they are limited to data in Euclidean space. In this paper, we propose a novel and mathematically principled framework that uses Riemannian metrics to reformulate DMPs such that the resulting formulation can operate with SPD data in the SPD manifold. Evaluation of the approach demonstrates that beneficial properties of DMPs such as change of the goal during operation apply also to the proposed formulation. Copyright © 2020, The Authors. All rights reserved.

关键词： Geometry