检索结果-内蒙古大学图书馆

SSRN 2024年

作者： Bahrami, Vahid Masouleh, Mehdi Tale Kalhor, Ahmad Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

This paper uses reinforcement learning techniques to introduce a novel approach to the controller design for a wide range of serial chain robots and cable-driven serial chain robots. The method addresses the regulation problem by designing a regulator and steady-state torques without requiring knowledge of the Coriolis vector, a complex part of the robot's dynamic model. The approach relies on Denavit-Hartenberg parameters to determine regulator and desired steady-state torques, using an actor-critic neural network for adaptation and a multi-layer perceptron neural network for desired steady-state torque. Stability is ensured through Lyapunov functions, with only the mass matrix and gravity vector required. The proposed control scheme offers adaptability, robustness, scalability, and optimization features compared to traditional controllers. Simulations on a 4-degrees-of-freedom cable-driven serial chain robot validate the proposed control method's effectiveness. A comparative benchmark is performed, demonstrating that the reinforcement learning controller outperforms other controllers in terms of regulation error, reducing it by at least 17.14\% compared to deep deterministic policy gradient and robust integral of the sign of the error methods. © 2024, The Authors. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Deep Learning-Based Imitation of human Actions for Autonomous Pick-and-Place Tasks

Deep Learning-Based Imitation of Human Actions for Autonomou...

引用

Iranian Conference on electrical engineering (ICEE)

作者： Anushe Saadati Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

ISBN: (数字)9798350376340

ISBN: (纸本)9798350376357

In robotics, attempts have been made to imitate movements through learning from demonstrations. There are two types of demonstration-based learning. Expert-based learning is expensive and time-consuming since it necessitates a paired observation-action procedure. Conversely, visual-based learning is more practical to collect and makes robots more adaptable to many kinds of tasks. But this problem still needs to be answered completely, and it's still a difficult assignment. This paper presents a comprehensive approach to replicating human actions in robots for object manipulation. As the main contribution of this paper, This research employs a deep learning methodology to allow robots to imitate human actions in grasping objects, utilizing human movement as a reference. A real dataset with 718 images and 142 videos has been collected in this paper. then The model is trained for object detection and object segmentation. To imitate human actions, a video of a person completing a task is recorded first. Following the subject's movements in the video allows for the extraction of the motion path. The robot then proceeds along the path that was extracted. The DPR impressively demonstrates the practical application, showcasing its capacity to imitate and execute human actions.

关键词： Deep learning Accuracy Tracking Video sequences Object detection Grasping Real-time systems

来源：评论

学校读者我要写书评

暂无评论

From Bricks to Bots: Automated Collision-Aware Sequence Planning for LEGO Reconstruction with a Two-Finger Gripper

From Bricks to Bots: Automated Collision-Aware Sequence Plan...

引用

International Conference on Control, Instrumentation, and Automation

作者： Arman Barghi Navid Pasiar Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

ISBN: (数字)9798331509972

ISBN: (纸本)9798331509989

An automated, vision-guided assembly method is proposed for reconstructing LEGO arrangements using a 3 DOF Delta Parallel robot with a two-finger gripper. Assembly sequence planning, an NP-complete problem, focuses on determining the optimal sequence of bricks for constructing a target structure. The complexity increases when robotic con-straints-such as avoiding collisions between the gripper and the partially constructed structure-are integrated into the planning. To address this, the Arrangement Planning Process is introduced, combining target arrangement identification with collision-aware sequence planning. This process includes precise detection techniques that merge instance segmentation with stud detection. Notable innovations include an assembly-by-disassembly algorithm for efficient planning and experimental techniques such as the Safe Approach and the Wiggling Technique for accurate brick placement. By implementing a homing method, positioning accuracy improved approximately twofold compared to previous work. The results demonstrate the feasibility and precision of the proposed method in automated LEGO reconstruction.

关键词： Instance segmentation Parallel robots Technological innovation Accuracy Three-dimensional displays Service robots Robustness Planning Grippers Assembly

来源：评论

学校读者我要写书评

暂无评论

Autonomous Rearrangement Planning Using Object Similarity Learning

Autonomous Rearrangement Planning Using Object Similarity Le...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Parmida Jabari Zeynab Ezzati Babi Hamed Ghasemi Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

ISBN: (数字)9798331529734

ISBN: (纸本)9798331529741

This study investigates the integration of artificial intelligence in robotics assembly systems, focusing on enhancing flexibility and adaptability in dynamic environments where the robot might not be given the precise objects of the intended construction. To this end, a developed approach is presented which is based on object detection, similarity assessment, and sequence planning for autonomous construction tasks. The system utilizes a YOLOv8 model for object detection and a Siamese network architecture, which facilitates object replacement. It processes two input images, one representing the target structure and another showing the available objects, extracting attributes such as bounding boxes, centers, and orientations. An exact object-matching step is performed before a two-stage similarity assessment. The two-stage similarity assessment prioritizes single-object matches and employs a multi-object similarity network only when necessary. This multi-stage approach is designed to maintain high similarity to the intended design while minimizing possible object shortcomings. During the construction phase, motion planning is performed using a 3-degrees-of-freedom Delta parallel robot, and a 2-finger gripper is utilized for robotics grasping. The results of this study, conducted on wooden bricks of varying shapes and sizes, demonstrate the effectiveness of the YOLOv8 model for object detection, achieving over 80% accuracy across 18 classes with a classification loss of approximately 0.25. Additionally, the Siamese network performed robustly, with a loss of less than 0.01 for both single and multiple object comparisons.

关键词： robotic assembly Parallel robots Automation Autonomous systems robot vision systems Warehousing Object detection Cameras Planning Grippers

来源：评论

学校读者我要写书评

暂无评论

Model-Free Dynamic Control of Five-bar Mechanism via Subchain Assembly and Reinforcement Learning

Model-Free Dynamic Control of Five-bar Mechanism via Subchai...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Soheil Sedghi Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

ISBN: (数字)9798331529734

ISBN: (纸本)9798331529741

Parallel robots are versatile but challenging to control due to their complex dynamics and the need for precise modeling. This research presents a novel approach to modular control of parallel robots, utilizing reinforcement learning to dynamically assemble and control complex arbitrary systems. By treating each sub chain of a parallel robot as an individual agent, the proposed method offers a flexible and adaptable framework for various control tasks. The study compares the learning performance of different RL algorithms and proposed decision modeling on the tracking control of a five-bar robot with a 55 cm link length, aiming to achieve an error of less than 5 cm in the PyBullet simulator. In single-agent reinforcement learning, three algorithms (DDPG, PPO, and SAC) were employed, with PPO and SAC effectively solving the problem. In multi-agent reinforcement learning, two algorithms (FACMAC and MAPPO) were utilized. While MAPPO successfully learned the problem, FACMAC exhibited difficulties in grasping the problem. The results obtained from comparing the proposed RL algorithms indicated that single-agent RL learns the control problem better and faster than multi-agent reinforcement learning. In short, the paper focuses on formulating an RL solution which can effectively control parallel robots and can potentially be used for pre-training subchains.

关键词： Training Parallel robots Adaptation models Sensitivity Mechatronics Heuristic algorithms Transfer learning Neural networks Training data Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Autonomous robotic Assembly and Sequence Planning Based on YOLOv8

Autonomous Robotic Assembly and Sequence Planning Based on Y...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Zeynab Ezzati Babi Navid Asadi Khomami Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran Human and Robot Interaction Laboratory School of Mechanical Engineering Tehran Iran

This study contributes to autonomous assembly by introducing a robotic building system designed to observe its workspace, plan building sequences, determine appropriate grasps and manipulator motions, and operate hardware in real-time. This study involves the creation of a wooden brick dataset and employs a YOLOv8 deep learning model for object segmentation in two implemented real-time algorithms, namely, Desired Construction Object Detection (DCOD) and robot Current Frame Object Detection (RCFOD). The DCOD algorithm is designed for the desired structure, while the RCFOD algorithm is utilized for detecting objects within the operational environment of the robot. By comparing the outputs of these two algorithms, sequences of objects for assembly are formulated, and the appropriate grasp positions are determined. Subsequently, the robot’s motion planning for both grasping and construction tasks is executed. Due to the limitations of using a single RGB gripper-mounted camera, various issues, such as failing to recognize items with the same cross-section but varying heights, have been addressed in this study. The results of this study indicate the effective learning of the YOLOv8 model, as shown by a loss value below 0.2 and an accuracy over 90% for 17 out of the total 18 classes. Practical implementation is done by a common industrial Delta parallel robot, a 2-finger gripper, and a simple RGB camera. Overview of practical implementation system is shown in Fig. 1. The proposed robotic assembly method demonstrates its effectiveness in autonomously constructing structures utilizing a reduced and cost-efficient configuration.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Dynamic Modeling of Double Segment Redundant Gough-Stewart Hybrid Manipulator based on the Principle of Virtual Work

Dynamic Modeling of Double Segment Redundant Gough-Stewart H...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Alireza Kamali Ardakani Hossein Akbari Parsa Namazian Mehdi Tale Masouleh Arash Bahrami Human and Robot Interaction Laboratory School of Mechanical Engineering University of Tehran Tehran Iran Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran School of Mechanical Engineering University of Tehran Tehran Iran

ISBN: (数字)9798331529734

ISBN: (纸本)9798331529741

This paper presents the dynamic modeling and analysis of a hyper-redundant manipulator with two Gough-Stewart platforms. Initially, the kinematic model of the manipulator is reviewed and used, since it acts as a base component for the subsequent dynamic modeling. Then, the virtual work approach is introduced to establish a foundation for the dynamic modeling of a single Gough-Stewart platform. This modeling for a single segment is used to form the equation of motion for both segments. However, the external forces applied to the lower Gough-Stewart robot are different from the upper one, since the force and masses in the upper platform act on the lower one. By considering these effects on the lower platform, the dynamic equations for the whole have been formed and presented. The accuracy of the proposed model is validated through a detailed comparison with simulations performed using a Simscape environment. Notable challenges in aligning the results between the model and simulations are discussed, highlighting the complexities involved in ensuring consistency across different platforms. The comparisons show an average error of 1.537 percent per actuator.

关键词： Accuracy Mechatronics robot kinematics Motion segmentation Dynamics Force Kinematics Mathematical models Complexity theory Manipulator dynamics

来源：评论

学校读者我要写书评

暂无评论

Real-Time Imitation of human Head Motions, Blinks and Emotions by Nao robot: A Closed-Loop Approach 11

Real-Time Imitation of Human Head Motions, Blinks and Emotio...

引用

11th RSI International Conference on robotics and Mechatronics, ICRoM 2023

作者： Rayati, Keyhan Feizi, Amirhossein Beigy, Alireza Shahverdi, Pourya Masouleh, Mehdi Tale Kalhor, Ahmad Louie, Wing-Yue Geoffrey University of Tehran Human and Robot Interaction Lab Electrical and Computer Engineering Tehran Iran Shahrood University of Technology Department of Electrical Engineering Shahrood Iran Oakland University Intelligent Robotics Laboratory Michigan United States

ISBN: (纸本)9798350308105

This paper introduces a novel approach for enabling real-time imitation of human head motion by a Nao robot, with a primary focus on elevating human-robot interactions. By using the robust capabilities of the MediaPipe as a computer vision library and the DeepFace as an emotion recognition library, this research endeavors to capture the subtleties of human head motion, including blink actions and emotional expressions, and seamlessly incorporate these indicators into the robot's responses. The result is a comprehensive framework which facilitates precise head imitation within human-robot interactions, utilizing a closed-loop approach that involves gathering real-time feedback from the robot's imitation performance. This feedback loop ensures a high degree of accuracy in modeling head motion, as evidenced by an impressive R2 score of 96.3 for pitch and 98.9 for yaw. Notably, the proposed approach holds promise in improving communication for children with autism, offering them a valuable tool for more effective interaction. In essence, proposed work explores the integration of real-time head imitation and real-time emotion recognition to enhance human-robot interactions, with potential benefits for individuals with unique communication needs. © 2023 IEEE.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

2D Skeleton-Based Keypoint Generation Method for Grasping Objects with Roughly Uniform Height Variation

2D Skeleton-Based Keypoint Generation Method for Grasping Ob...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Ali Sabzejou Mehdi Tale Masouleh Ahmad Kalhor School of Electrical and Computer Engineering Department Human and Robot Interaction Laboratory University of Tehran

robotic grasping poses a fundamental challenge in robotics, particularly when dealing with unknown objects. This skill has always been a focal point in robotics research due to its fundamental importance and inherent complexity. Recent advances utilize state-of-the-art learning techniques, including deep learning and deep reinforcement learning, presenting their unique set of challenges. Training these models requires extensive data and computational resources, with the generalization to unknown objects presenting a significant obstacle for robots. This paper presents a comprehensive approach to unknown object robotic grasping, with a specific focus on top-down grasping actions. The process encompasses image preprocessing, the application of the Straight Skeleton (StSkel) method, and the systematic generation of grasp keypoints when applied to a selection of objects from the Dex-Net dataset. During the evaluation, the analysis incorporated crucial metrics, including the number of detected grasp keypoints for each object, the count of successfully generated grasp pairs, the overall success rate, and the success rate when considering the top 5 ranked grasp pairs. One of the notable strengths of this approach lies in its adaptability to a wide array of objects, ranging from simple shapes to complex ones. The proposed approach in this paper paves the way for automatically labeling grasping datasets, providing a valuable asset for developing auto-generating Deep-RL models. The StSkel method effectively captures essential structural information, making it suitable for real-world applications involving diverse objects. The experimental outcomes affirm the robustness and adaptability of the approach, as it most often achieved a high success rate in generating viable pairs of grasping points for most tested objects. Even complex objects yield impressive results, demonstrating the method’s potential for real-world applications in grasping unknown objects.

关键词：

来源：评论

学校读者我要写书评

暂无评论

HUGGA: human-like Grasp Generation with Gripper’s Approach State Using Deep Learning

HUGGA: Human-like Grasp Generation with Gripper’s Approach ...

引用

RSI/ISM International Conference on robotics and Mechatronics (ICROM)

作者： Elnaz Balazadeh Mehdi Tale Masouleh Ahmad Kalhor Human and Robot Interaction Laboratory University of Tehran Tehran Iran Human and Robot Interaction Laboratory School of Electrical and Computer Engineering University of Tehran Tehran Iran

In this study, an attempt has been made to develop a new method for producing a successful grasp based on human behavior, by considering the gripper’s approach state in the final grasp of an object. The dataset collected for the purpose of this paper consists of 14 different daily usage objects with different orientations (In total 2526 images) which a human approaches and grasps them. Data collection has been done based on a protocol shown in TableI to cover all achievable states. From Fig1, it can be seen that the dataset includes one image from a single object without the user’s hand. The second image contains the human hand approach and its configuration regarding the object which is subject to be grasped and the third image represents the object which is grasped by the human hand and thus evidently contains both object and human hand grasp configuration. It is important to acknowledge that providing a static hand image does not suffice to facilitate the acquisition of human-like grasping capabilities by robot grippers, so, a mapping from hand fingers to a two-finger gripper should be implemented to fill this gap. In the collected dataset, all grasps are performed by the index finger and thumb to make it possible to map the coordinates of each finger on a gripper finger. Using MediaPipe, which operates a single shot detection network, with only an RGB image containing a human hand the index finger and thumb landmarks were extracted. Finally, a detailed human approach and grasp information were available, and with the collected data, a two-input and one-output deep neural network was trained. The designed model produced a grasp with a human pattern by receiving only the RGB image of the object and the approach vector to the object. Ultimately, through testing and evaluation based on existing criteria, the trained network successfully produced a grasp with 61% accuracy in grasp rectangle overlap, 97% in grasp angle and overall 60.5% met two established grasp crite

关键词：

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：