检索结果-内蒙古大学图书馆

Adaptive feedback compensation control method for bipedal robot walking under continuous external disturbances

Acta Mechanica Sinica 2024年第12期40卷 29-39页

作者： Zijing Li Jinlin Zhang Mengyue Lu Wanchao Chi Chong Zhang Shenghao Zhang Yuzhen Liu Chunbiao Gan State Key Laboratory of Fluid Power and Mechatronic Systems School of Mechanical EngineeringZhejiang UniversityHangzhou 310058China Tencent Robotics X Tencent Binhai BuildingShenzhen 518054China

In the past few decades,people have been trying to address the issue of walking instability in bipedal robots in uncertain ***,most control methods currently have still failed to achieve robust walking of bipedal robots under uncertain *** research mostly focuses on motion control methods for robots on uneven terrain and under sudden impact forces,with little consideration for the problem of continuous and intense external force disturbances in uncertain *** response to this issue,a disturbance-robust control method based on adaptive feedback compensation is ***,based on the Lagrangian method,the dynamic model of a bipedal robot under different types of external force disturbances was ***,through dynamic analysis,it was observed that classical control methods based on hybrid zero dynamics failed to consider the continuous and significant external force disturbances in uncertain ***,an adaptive feedback compensation controller was designed,and an adaptive parameter adjustment optimization algorithm was proposed based on walking constraints to achieve stable walking of bipedal robots under different external force ***,in numerical simulation experiments,comparative analysis revealed that using only a controller based on hybrid zero dynamics was insufficient to converge the motion of a planar five-link bipedal robot subjected to periodic forces or bounded noise disturbances to a stable *** contrast,in the adaptive feedback compensation control method,the use of an adaptive parameter adjustment optimization algorithm to generate time-varying control parameters successfully achieved stable walking of the robot under these *** indicates the effectiveness of the adaptive parameter adjustment algorithm and the robustness of the adaptive feedback compensation control method.

关键词： Bipedal robot External disturbance Walking stability Adaptive feedback compensation Anti-disturbance control

来源：评论

学校读者我要写书评

暂无评论

Single Modality and Joint Fusion for Emotion Recognition on RAVDESS Dataset

引用

SN Computer Science 2024年第6期5卷 669页

作者： Haddad, Syrine Daassi, Olfa Belghith, Safya Laboratory of Robotics Informatics and Complex Systems (RISC Lab-LR16ES07) Tunis Tunisia National Engineering School of Tunis University of Tunis El manar Tunis 1002 Tunisia National Engineering School of Carthage University of Carthage Carthage 2035 Tunisia

Over the last decade, emotion recognition has become a widely researched area worth considering in any project related to affective computing. Due to the limitless applications of this new discipline, the development of emotion recognition systems has emerged as a lucrative opportunity in the corporate sector. Emotion recognition can be detected in many ways, such as face, speech, text, gestures, etc. Hence, this article presents a comprehensive survey of emotion recognition methods, focusing on three key modalities: visual, speech, and audio-visual. Moreover, a novel approach for each modality is proposed, and a comparative analysis is conducted with existing methods to assess their effectiveness. The results of this study revealed that face emotion recognition on the RAVDESS dataset can achieve good results compared to speech emotion recognition, which was the opposite case in previous literature. Combined modalities were found to likely achieve even better results with an accuracy of 89% compared to previous works. Finally, this article concludes by offering new insights and suggestions for future research. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.

关键词： Computational models Face emotion recognition Multi-modal emotion recognition RAVDESS dataset Speech emotion recognition

来源：评论

学校读者我要写书评

暂无评论

Active Camera View Planning Based on Deep Reinforcement Learning in Indoor Exploration 42

Active Camera View Planning Based on Deep Reinforcement Lear...

引用

42nd Chinese Control Conference, CCC 2023

作者： Wang, Hao-Yu Meng, Qing-Hao Dai, Xu-Yang Institute of Robotics and Autonomous Systems School of Electrical and Information Engineering Tianjin University Tianjin Key Laboratory of Intelligent Unmanned Swarm Technology and System Tianjin300072 China

ISBN: (纸本)9789887581543

In the last decades, significant progress has been made in the field of active vSLAM (visual simultaneous localization and mapping). However, the majority of active vSLAM methods suffer from the problem of coupling the camera's viewpoint to the robot's motion, which limits the initiative during the exploration of unknown environments. To address this issue, this study proposes a camera view planning method based on deep reinforcement learning (DRL), which makes the camera rotate independently. First, feature encoding is performed on the reconstructed navigation map using a multi-layer convolutional neural network (CNN). Second, the DRL-based camera view planning method is trained through the proximal policy optimization (PPO) algorithm. Last, a significant number of tests is performed on a public dataset of indoor environments. The simulation results show that the proposed camera view planning method improves the exploration coverage by an average of 4% (33.86 m2 for traditional vs. 35.21 m2 for the suggested method). Meanwhile, the kind of semantic information newly observed in each step increases by 21% (2.9 for traditional vs. 3.5 for the proposed method) on average, indicating the improvement in the efficiency of information acquisition. © 2023 Technical Committee on Control Theory, Chinese Association of Automation.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Pegasus Simulator: An Isaac Sim Framework for Multiple Aerial Vehicles Simulation

Pegasus Simulator: An Isaac Sim Framework for Multiple Aeria...

引用

International Conference on Unmanned Aircraft systems (ICUAS)

作者： Marcelo Jacinto João Pinto Jay Patrikar John Keller Rita Cunha Sebastian Scherer António Pascoal Laboratory of Robotics and Engineering Systems ISR/IST University of Lisbon Portugal Robotics Institute Carnegie Mellon University USA

ISBN: (数字)9798350357882

ISBN: (纸本)9798350357899

Developing and testing novel control and motion planning algorithms for aerial vehicles can be a challenging task, with the robotics community relying more than ever on 3D simulation technologies to evaluate the performance of new algorithms in a variety of conditions and environments. In this work, we introduce the Pegasus Simulator, a modular framework implemented as an NVIDIA ® Isaac Sim extension that enables real-time simulation of multiple multirotor vehicles in photo-realistic environments, while providing out-of-the-box integration with the widely adopted PX4-Autopilot and ROS2 through its modular implementation and intuitive graphical user interface. To demonstrate some of its capabilities, a nonlinear controller was implemented and simulation results for two drones performing aggressive flight maneuvers are presented. Code and documentation for this framework are also provided as supplementary material.

关键词： Three-dimensional displays Simulation Robot sensing systems Real-time systems Sensor systems Sensors Task analysis

来源：评论

学校读者我要写书评

暂无评论

Accelerated Informed RRT*: Fast and Asymptotically Path Planning Method Combined with RRT*-Connect and APF 16th

Accelerated Informed RRT*: Fast and Asymptotically Path Pla...

引用

16th International Conference on Intelligent robotics and Applications, ICIRA 2023

作者： Tu, Zhixin Zhuang, Wenbing Leng, Yuquan Fu, Chenglong Department of Mechanical and Energy Engineering Southern University of Science and Technology Shenzhen518055 China Shenzhen Key Laboratory of Biomimetic Robotics and Intelligent Systems and Guangdong Provincial Key Laboratory of Human-Augmentation and Rehabilitation Robotics in Universities Southern University of Science and Technology Shenzhen518055 China

ISBN: (纸本)9789819965007

In recent years, path planning algorithms have played a crucial role in addressing complex navigation problems in various domains, including robotics, autonomous vehicles, and virtual simulations. This abstract introduces a improved path planning algorithm called Informed RRT*-connect based on APF, which combines the strengths of the fast bidirectional rapidly-exploring random tree (RRT-connect) algorithm and the informed RRT* algorithm. The proposed algorithm aims to efficiently find collision-free paths with less iterations and time while minimizing the path length. Unlike traditional RRT-based algorithms, Informed RRT*-connect based on Artificial Potential Fields (APF) incorporates a bidirectional connection and rewiring of a new sampling point to explore the search space. This enables the algorithm to connect both the start and goal nodes more effectively and quickly to find a initial solution, reducing the search time and provide a better initial heuristics sapling for the next optimal steps. Furthermore, Informed RRT*-connect introduces an informed sampling strategy that biases the sampling towards areas of the configuration space likely to yield better paths. This approach significantly reduces the exploration time to find a path and enhances the ability to discover optimal paths efficiently. To evaluate the effectiveness of the Informed RRT*-connect algorithm, we conducted the simulation experiments on two different experiment protocol. The results demonstrate that our approach outperforms existing state-of-the-art algorithms in terms of both planning efficiency and solution optimality. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023.

关键词： Motion planning

来源：评论

学校读者我要写书评

暂无评论

Motion Capture and Gait Analysis of Ants with Leg Injuries 42

Motion Capture and Gait Analysis of Ants with Leg Injuries

引用

42nd Chinese Control Conference, CCC 2023

作者： Zeng, Ming Li, Yuanhao Meng, Chang Zhong, Shutong Wang, Zhijing Zhao, Feng Institute of Robotics and Automation Systems School of Electrical and Information Engineering Tianjin University Tianjin Key Laboratory of Intelligent Unmanned Swarm Technology and System Tianjin300072 China

ISBN: (纸本)9789887581543

Many insects in nature, such as ants, recover quickly after sustaining leg injuries. If this capability could be imparted to multi-legged robots, it would drastically enhance their ability to resist destruction and adapt to the environment. In this paper, the Camponotus Japonicus is used as the research object, and its locomotive patterns are studied under different leg-loss conditions. To facilitate this, a high-speed acquisition system, with a dual-view (bottom view and side view) setup, is designed to analyze the ant's linear gaits. Subsequently, the pivotal gait characteristics of the injured ants are extracted by keyframe annotation technology, including the gait phase diagram, duty factor, angle before support, angle before swing, and leg swing range. Lastly, according to the obtained gait characteristics, the general and specific laws of gait adjustment in the absence of one or two legs are systematically analyzed, which provides the basis for parametric settings and strategy optimization of robotic injured gait self-adjustment. © 2023 Technical Committee on Control Theory, Chinese Association of Automation.

关键词： Gait analysis

来源：评论

学校读者我要写书评

暂无评论

Improving Collision-Free Success Rate For Object Goal Visual Navigation Via Two-Stage Training With Collision Prediction

arXiv

引用

arXiv 2025年

作者： Lian, Shiwei Zhang, Feitian Robotics and Control Laboratory Department of Advanced Manufacturing and Robotics College of Engineering State Key Laboratory of Turbulence and Complex Systems Peking University Beijing100871 China

The object goal visual navigation is the task of navigating to a specific target object using egocentric visual observations. Recent end-to-end navigation models based on deep reinforcement learning have achieved remarkable performance in finding and reaching target objects. However, the collision problem of these models during navigation remains unresolved, since the collision is typically neglected when evaluating the success. Although incorporating a negative reward for collision during training appears straightforward, it results in a more conservative policy, thereby limiting the agent’s ability to reach targets. In addition, many of these models utilize only RGB observations, further increasing the difficulty of collision avoidance without depth information. To address these limitations, a new concept—collision-free success is introduced to evaluate the ability of navigation models to find a collision-free path towards the target object. A two-stage training method with collision prediction is proposed to improve the collision-free success rate of the existing navigation models using RGB observations. In the first training stage, the collision prediction module supervises the agent’s collision states during exploration to learn to predict the possible collision. In the second stage, leveraging the trained collision prediction, the agent learns to navigate to the target without collision. The experimental results in the AI2-THOR environment demonstrate that the proposed method greatly improves the collision-free success rate of different navigation models and outperforms other comparable collision-avoidance methods. Copyright © 2025, The Authors. All rights reserved.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Feature Space Augmentation and Old Class Space Preservation for Class Incremental Learning 42

Feature Space Augmentation and Old Class Space Preservation ...

引用

42nd Chinese Control Conference, CCC 2023

作者： Zeng, Ming Li, Siying Liu, Jiazhe Li, Yuanhao Institute of Robotics and Automation Systems School of Electrical and Information Engineering Tianjin Key Laboratory of Intelligent Unmanned Swarm Technology and System Tianjin University Tianjin300072 China

ISBN: (纸本)9789887581543

The ability of continuously learning new knowledge without forgetting old ones is crucial to adapt to an ever-changing world. This scenario becomes more challenging when the previous data are not available. Current class incremental learning (CIL) approaches tend to incorporate new classes with backward compatibility, i.e., maintaining discriminability of old ones. By contrast, we focus on the extensibility and compatibility for future new classes in the early stage. Our proposed method achieves this by augmenting feature space. In detail, a large number of pseudo-new classes are generated via real example mixture and then train the initial spatially augmented model using pseudo-new classes and the base classes. Besides, considering the fact that it is impossible to maintain backward compatibility with only one prototype for each old class, the updated model also needs to preserve old class space in incremental stages. We employ a certain perturbation to the old class prototypes to effectively avoid old class space from being over-squeezed by the samples of new classes. The experiments on CIFAR-100 and ImageNet-Subset (100 classes) demonstrate that our method substantially reduces the overlap of old and new classes, outperforming state-of-the-art various baselines. © 2023 Technical Committee on Control Theory, Chinese Association of Automation.

关键词： Catastrophic forgetting Class incremental learning Feature space augmentation Old class space preservation Pseudo-new classes

来源：评论

学校读者我要写书评

暂无评论

Application Research of Landmarks in Path Planning Algorithms of Mobile Robots 2

Application Research of Landmarks in Path Planning Algorithm...

引用

2nd International Conference on Digital Society and Intelligent systems, DSInS 2022

作者： Fan, Linghao Liu, Xiangde Zhang, Yi Chongqing University of Posts and Telecommunications Intelligent Systems and Robotics Laboratory Chongqing China Chongqing Information Accessibility and Service Robot Engineering Technology Research Center Chongqing China

ISBN: (纸本)9781510663244

A way of incorporating landmark information matrix QR to affect the global path planning algorithm is proposed for the use of landmarks in robot localization and navigation. It is also investigated how this method affects the pace at which the original algorithm solves problems. The reward and punishment items in this technique include dangerous locations with lots of obstacles around them, landmarks with location information connected to the ground, and the two combined. Three representative common robot path planning algorithms are also imposed as influences. Through MATLAB simulation experiment, find the optimal path planning algorithm that is appropriate for applying reward and punishment by comparing the quality of the robot's driving path and the algorithm solution speed before and after the reward and punishment. According to the simulation results, the reward and punishment approach is practical and efficient. The heuristic search algorithm is better suited for reward and punishment, and both its path quality and speed of solution have greatly increased. © 2023 SPIE.

关键词： Genetic algorithms

来源：评论

学校读者我要写书评

暂无评论

Trajectory Control of Multi-Axis Robotic Arms Based on Physics-Informed Neural Networks and Nonlinear Model Predictive Control

Trajectory Control of Multi-Axis Robotic Arms Based on Physi...

引用

2024 China Automation Congress, CAC 2024

作者： Wang, Yiqing Xia, Changqing Jin, Xi Xu, Chi Zhang, Yanzhu School of Automation and Electrical Engineering Shenyang Ligong University Shenyang China State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China Key Laboratory of Networked Control Systems Chinese Academy of Sciences Shenyang China

ISBN: (纸本)9798350368604

Multi-axis robotic arms are extensively utilized in intelligent manufacturing scenarios, with trajectory control in flexible scenarios constituting a primary challenge. Physics-Informed Neural Networks (PINNs) represent advanced methods that integrate physical laws with data-driven approaches. Despite their outstanding performance in integrating theory and data, research on their application for controlling robotic arms in complex scenarios remains limited. This paper establishes a nonlinear dynamic model based on the physical relationships between the axes of robotic arms. In the absence of prior data, Sobol sequence random sampling is employed to generate data, which are subsequently trained using PINNs. Considering system noise, the Extended Kalman Filter (EKF) is utilized to predict the next state in noisy environments, and Nonlinear Model Predictive Control (NMPC) is implemented to control the robotic arm for trajectory tracking, achieving real-time control. Simulation results demonstrate that the proposed Discrete Physics-Informed Predictive Control (DPIPC) method exhibits smaller position and velocity errors with less fluctuation compared to the method in the reference, indicating superior control capabilities. © 2024 IEEE.

关键词： Extended Kalman filters

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：