检索结果-内蒙古大学图书馆

SGANVO: Unsupervised deep Visual Odometry and Depth Estimation With Stacked Generative Adversarial Networks

IEEE robotics AND automation LETTERS 2019年第4期4卷 4431-4437页

作者： Feng, Tuo Gu, Dongbing Univ Essex Sch Comp Sci & Elect Engn Colchester CO4 3SQ Essex England

Recently end-to-end unsupervised deep learning methods have demonstrated an impressive performance for visual depth and ego-motion estimation tasks. These data-based learning methods do not rely on the same limiting assumptions that geometry-based methods do. The encoder-decoder network has been widely used in the depth estimation and the RCNN has brought significant improvements in the ego-motion estimation. Furthermore, the latest use of generative adversarial nets (GANs) in depth and ego-motion estimation has demonstrated that the estimation could be further improved by generating pictures in the game learning process. This paper proposes a novel unsupervised network system for visual depth and ego-motion estimation-stacked generative adversarial network. It consists of a stack of GAN layers, of which the lowest layer estimates the depth and ego-motion while the higher layers estimate the spatial features. It can also capture the temporal dynamic due to the use of a recurrent representation across the layers. We select the most commonly used KITTI data set for evaluation. The evaluation results show that our proposed method can produce better or comparable results in depth and ego-motion estimation.

关键词： Estimation Generators Image reconstruction Generative adversarial networks Visual odometry Training Gallium nitride deep learning in robotics and automation SLAM Localization Mapping

来源：评论

学校读者我要写书评

暂无评论

Robot-Assisted Training in Laparoscopy Using deep Reinforcement learning

引用

IEEE robotics AND automation LETTERS 2019年第2期4卷 485-492页

作者： Tan, Xiaoyu Chng, Chin-Boon Su, Ye Lim, Kah-Bin Chui, Chee-Kong Natl Univ Singapore Dept Mech Engn Singapore 119077 Singapore

Minimally invasive surgery (MIS) is increasingly becoming a vital method of reducing surgical trauma and significantly improving postoperative recovery. However, skillful handling of surgical instruments used in MIS, especially for laparoscopy, requires a long period of training and depends highly on the experience of surgeons. This letter presents a new robot-assisted surgical training system which is designed to improve the practical skills of surgeons through intrapractice feedback and demonstration from both human experts and reinforcement learning (RL) agents. This system utilizes proximal policy optimization to learn the control policy in simulation. Subsequently, a generative adversarial imitation learning agent is trained based on both expert demonstrations and learned policies in simulation. This agent then generates demonstration policies on the robot-assisted device for trainees and produces feedback scores during practice. To further acquire surgical tools coordinates and encourage self-oriented practice, a mask region-based convolution neural network is trained to perform the semantic segmentation of surgical tools and targets. To the best of our knowledge, this system is the first robot-assisted laparoscopy training system which utilizes actual surgical tools and leverages deep reinforcement learning to provide demonstration training from both human expert perspectives and RL criterion.

关键词： Surgical robotics laparoscopy learning from demonstration deep learning in robotics and automation AI-based methods

来源：评论

学校读者我要写书评

暂无评论

Convolutional Autoencoder for Feature Extraction in Tactile Sensing

引用

IEEE robotics AND automation LETTERS 2019年第4期4卷 3671-3678页

作者： Polic, Marsela Krajacic, Ivona Lepora, Nathan Orsag, Matko Univ Zagreb Fac Elect Engn & Comp Lab Robot & Intelligent Control Syst Zagreb 10000 Croatia Univ Bristol Dept Engn Math Bristol BS8 1UB Avon England Univ Bristol Bristol Robot Lab Bristol BS8 1UB Avon England

A common approach in the field of tactile robotics is the development of a new perception algorithm for each new application of existing hardware solutions. In this letter, we present a method of dimensionality reduction of an optical-based tactile sensor image output using a convolutional neural network encoder structure. Instead of using various complex perception algorithms, and/ or manually choosing task-specific data features, this unsupervised feature extraction method allows simultaneous online deployment of multiple simple perception algorithms on a common set of black-box features. The method is validated on a set of benchmarking use cases. Contact object shape, edge position, orientation, and indentation depth are estimated using shallowneural networks and machine learning models. Furthermore, a contact force estimator is trained, affirming that the extracted features contain sufficient information on both spatial and mechanical characteristics of the manipulated object.

关键词： Force and tactile sensing deep learning in robotics and automation soft sensors and actuators perception for grasping and manipulation

来源：评论

学校读者我要写书评

暂无评论

Efficient collection and automatic annotation of real-world object images by taking advantage of post-diminished multiple visual markers

引用

ADVANCED robotics 2019年第24期33卷 1264-1280页

作者： Kiyokawa, Takuya Tomochika, Keita Takamatsu, Jun Ogasawara, Tsukasa Nara Inst Sci & Technol Div Informat Sci Nara Japan

To collect a human-annotated dataset for training deep convolutional neural networks is a very time-consuming and laborious process. To reduce this burden, we previously proposed an automated annotation by placing one visual marker above the detection target object in the training phase. However, in this approach, occasionally the marker hides the object surface. To avoid this issue, we propose placing a pedestal with multiple markers at the bottom of the object. If we use multiple markers, the object can be annotated even when the object hides some of the markers. Besides that, the simple modification of placing the markers on the bottom allows the use of simple background masking to avoid the neural network learning the remaining markers in the training image as a feature of the object. Background masking can completely remove the markers during the training process. Experiments showed the proposed vision system using our automatic object annotation outperformed the vision system using manual annotation in terms of object detection, orientation estimation, and 2D position estimation while reducing the time required for dataset collection from 16.1 hours to 7.30 hours.

关键词： Automatic annotation computer vision for automation deep learning in robotics and automation object detection and pose estimation

来源：评论

学校读者我要写书评

暂无评论

Low-Level Control of a Quadrotor With deep Model-Based Reinforcement learning

引用

IEEE robotics AND automation LETTERS 2019年第4期4卷 4224-4230页

作者： Lambert, Nathan O. Drewe, Daniel S. Yaconelli, Joseph Levine, Sergey Calandra, Roberto Pister, Kristofer S. J. Univ Calif Berkeley Dept Elect Engn & Comp Sci Berkeley CA 94720 USA Univ Oregon Eugene OR 97403 USA Facebook AI Res Menlo Pk CA 94025 USA

Designing effective low-level robot controllers often entail platform-specific implementations that require manual heuristic parameter tuning, significant system knowledge, or long design times. With the rising number of robotic and mechatronic systems deployed across areas ranging from industrial automation to intelligent toys, the need for a general approach to generating low-level controllers is increasing. To address the challenge of rapidly generating low-level controllers, we argue for using model-based reinforcement learning (MBRL) trained on relatively small amounts of automatically generated (i.e., without system simulation) data. In this letter, we explore the capabilities of MBRL on a Crazyflie centimeter-scale quadrotor with rapid dynamics to predict and control at <= 50 Hz. To our knowledge, this is the first use of MBRL for controlled hover of a quadrotor using only on-board sensors, direct motor input signals, and no initial dynamics knowledge. Our controller leverages rapid simulation of a neural network forward dynamics model on a graphic processing unit enabled base station, which then transmits the best current action to the quadrotor firmware via radio. In our experiments, the quadrotor achieved hovering capability of up to 6 s with 3 min of experimental training data.

关键词： deep learning in robotics and automation aerial systems: mechanics and control

来源：评论

学校读者我要写书评

暂无评论

learning Navigation Behaviors End-to-End With AutoRL

引用

IEEE robotics AND automation LETTERS 2019年第2期4卷 2007-2014页

作者： Chiang, Hao-Tien Lewis Faust, Aleksandra Fiser, Marek Francis, Anthony Google AI Google Robot Mountain View CA 94043 USA

We learn end-to-end point-to-point and pathfollowing navigation behaviors that avoid moving obstacles. These policies receive noisy lidar observations and output robot linear and angular velocities. The policies are trained in small, static environments with AutoRL, an evolutionary automation layer around reinforcement learning (RL) that searches for a deep RL reward and neural network architecture with large-scale hyper-parameter optimization. AutoRL first finds a reward that maximizes task completion and then finds a neural network architecture that maximizes the cumulative of the found reward. Empirical evaluations, both in simulation and on-robot, show that AutoRL policies do not suffer from the catastrophic forgetfulness that plagues many other deep reinforcement learning algorithms, generalize to new environments and moving obstacles, are robust to sensor, actuator, and localization noise, and can serve as robust building blocks for larger navigation tasks. Our path-following and point-to-point policies are, respectively, 23% and 26% more successful than comparison methods across new environments.

关键词： Autonomous agents collision avoidance deep learning in robotics and automation motion and path planning

来源：评论

学校读者我要写书评

暂无评论

Sliding-Window Temporal Attention Based deep learning System for Robust Sensor Modality Fusion for UGV Navigation

引用

IEEE robotics AND automation LETTERS 2019年第4期4卷 4216-4223页

作者： Unlu, Halil Utku Patel, Naman Krishnamurthy, Prashanth Khorrami, Farshad NYU Tandon Sch Engn Dept Elect & Comp Engn Controls Robot Res Lab Brooklyn NY 11201 USA

We propose a novel temporal attention based neural network architecture for robotics tasks that involve fusion of time series of sensor data, and evaluate the performance improvements in the context of autonomous navigation of unmanned ground vehicles (UGVs) in uncertain environments. The architecture generates feature vectors by fusing raw pixel and depth values collected by camera(s) and LiDAR(s), stores a history of the generated feature vectors, and incorporates the temporally attended history with current features to predict a steering command. The experimental studies show the robust performance in unknown and cluttered environments. Furthermore, the temporal attention is resilient to noise, bias, blur, and occlusions in the sensor signals. We trained the network on indoor corridor datasets (that will be publicly released) from our UGV. The datasets have LiDAR depth measurements, camera images, and human tele-operation commands.

关键词： Autonomous vehicle navigation sensor fusion deep learning in robotics and automation

来源：评论

学校读者我要写书评

暂无评论

RL-RRT: Kinodynamic Motion Planning via learning Reachability Estimators From RL Policies

引用

IEEE robotics AND automation LETTERS 2019年第4期4卷 4298-4305页

作者： Chiang, Hao-Tien Lewis Hsu, Jasmine Fiser, Marek Tapia, Lydia Faust, Aleksandra Google AI Robot Google Mountain View CA 94043 USA Univ New Mexico Dept Comp Sci Albuquerque NM 87131 USA

This letter addresses two challenges facing sampling-based kinodynamic motion planning: a way to identify good candidate states for local transitions and the subsequent computationally intractable steering between these candidate states. Through the combination of sampling-based planning, a Rapidly Exploring Randomized Tree (RRT) and an efficient kinodynamic motion planner through machine learning, we propose an efficient solution to long-range planning for kinodynamic motion planning. First, we use deep reinforcement learning to learn an obstacle-avoiding policy that maps a robot's sensor observations to actions, which is used as a local planner during planning and as a controller during execution. Second, we train a reachability estimator in a supervised manner, which predicts the RL policy's time to reach a state in the presence of obstacles. Lastly, we introduce RL-RRT that uses the RL policy as a local planner, and the reachability estimator as the distance function to bias tree-growth towards promising regions. We evaluate our method on three kinodynamic systems, including physical robot experiments. Results across all three robots tested indicate that RL-RRT outperforms state of the art kinodynamic planners in efficiency, and also provides a shorter path finish time than a steering function free method. The learned local planner policy and accompanying reachability estimator demonstrate transferability to the previously unseen experimental environments, making RL-RRT fast because the expensive computations are replaced with simple neural network inference.

关键词： Motion and path planning learning and adaptive systems deep learning in robotics and automation

来源：评论

学校读者我要写书评

暂无评论

On-Policy Dataset Synthesis for learning Robot Grasping Policies Using Fully Convolutional deep Networks

引用

IEEE robotics AND automation LETTERS 2019年第2期4卷 1357-1364页

作者： Satish, Vishal Mahler, Jeffrey Goldberg, Ken Univ Calif Berkeley Dept Elect Engn & Comp Sci Berkeley CA 94705 USA Univ Calif Berkeley Dept Ind Operat & Engn Res Berkeley CA 94705 USA

Rapid and reliable robot grasping for a diverse set of objects has applications from warehouse automation to home de-cluttering. One promising approach is to learn deep policies from synthetic training datasets of point clouds, grasps, and rewards sampled using analytic models with stochastic noise models for domain randomization. In this letter, we explore how the distribution of synthetic training examples affects the rate and reliability of the learned robot policy. We propose a synthetic data sampling distribution that combines grasps sampled from the policy action set with guiding samples from a robust grasping supervisor that has full state knowledge. We use this to train a robot policy based on a fully convolutional network architecture that evaluates millions of grasp candidates in 4-DOF (3-D position and planar orientation). Physical robot experiments suggest that a policy based on fully convolutional grasp quality CNNs (FC-GQ-CNNs) can plan grasps in 0.625 s, considering 5000x more grasps than our prior policy based on iterative grasp sampling and evaluation. This computational efficiency improves rate and reliability, achieving 296 mean picks per hour (MPPH) compared to 250 MPPH for iterative policies. Sensitivity experiments explore the effect of supervisor guidance level and granularity of the policy action space. Code, datasets, videos, and supplementary material can be found at http://***/fcgqcnn.

关键词： Grasping deep learning in robotics and automation

来源：评论

学校读者我要写书评

暂无评论

Multi-View Incremental Segmentation of 3-D Point Clouds for Mobile Robots

引用

IEEE robotics AND automation LETTERS 2019年第2期4卷 1240-1246页

作者： Chen, Jingdao Cho, Yong Kwon Kira, Zsolt Georgia Inst Technol Inst Robot & Intelligent Machines Atlanta GA 30332 USA Georgia Inst Technol Sch Civil & Environm Engn Atlanta GA 30332 USA

Mobile robots need to create high-definition three-dimensional (3-D) maps of the environment for applications such as remote surveillance and infrastructure mapping. Accurate semantic processing of the acquired 3-D point cloud is critical for allowing the robot to obtain a high-level understanding of the surrounding objects and perform context-aware decision making. Existing techniques for point cloud semantic segmentation are mostly applied on a single frame or offline basis, with no way to integrate the segmentation results over time. This letter proposes an online method for mobile robots to incrementally build a semantically rich 3-D point cloud of the environment. The proposed deep neural network, MCPNet, is trained to predict class labels and object instance labels for each point in the scanned point cloud in an incremental fashion. A multi-view context pooling (MCP) operator is used to combine point features obtained from multiple viewpoints to improve the classification accuracy. The proposed architecture was trained and evaluated on ray-traced scans derived from the Stanford 3-D Indoor Spaces dataset. Results show that the proposed approach led to 15% improvement in pointwise accuracy and 7% improvement in normalized mutual information compared to the next best online method, with only a 6% drop in accuracy compared to the PointNet-based offline approach.

关键词： deep learning in robotics and automation object detection segmentation and categorization semantic scene understanding

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：