Pub Date : 2024-09-13DOI: 10.1109/LRA.2024.3461550
Jayant Unde;Jacinto Colan;Yasuhisa Hasegawa
This letter presents a novel approach to fabric manipulation through the development and optimization of a single-actuator-driven roller gripper. Focused on addressing the challenges inherent in handling fabrics with diverse thicknesses and materials, our gripper employs a passive adaptable mechanism driven by springs, enabling effective manipulation of fabrics ranging from 0.1 mm to 2.25 mm in thickness. We analyze gripper-fabric interaction forces to identify the parameters that influence successful grasping. We then optimize the gripper's normal forces and the roller's tangential force using the proposed model. Systematic evaluations demonstrated the gripper's capability to separate individual layers from fabric stacks, achieving a 94.9% success rate across multiple fabric types. Overall, this research offers a compact, cost-effective solution with broad applicability in diverse industrial automation contexts, providing valuable insights for advancing robotic fabric handling systems.
{"title":"Design, Modeling, and Experimental Verification of Passively Adaptable Roller Gripper for Separating Stacked Fabric","authors":"Jayant Unde;Jacinto Colan;Yasuhisa Hasegawa","doi":"10.1109/LRA.2024.3461550","DOIUrl":"https://doi.org/10.1109/LRA.2024.3461550","url":null,"abstract":"This letter presents a novel approach to fabric manipulation through the development and optimization of a single-actuator-driven roller gripper. Focused on addressing the challenges inherent in handling fabrics with diverse thicknesses and materials, our gripper employs a passive adaptable mechanism driven by springs, enabling effective manipulation of fabrics ranging from 0.1 mm to 2.25 mm in thickness. We analyze gripper-fabric interaction forces to identify the parameters that influence successful grasping. We then optimize the gripper's normal forces and the roller's tangential force using the proposed model. Systematic evaluations demonstrated the gripper's capability to separate individual layers from fabric stacks, achieving a 94.9% success rate across multiple fabric types. Overall, this research offers a compact, cost-effective solution with broad applicability in diverse industrial automation contexts, providing valuable insights for advancing robotic fabric handling systems.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10680377","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142274956","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-11DOI: 10.1109/LRA.2024.3458807
Yifang Zhang;Arash Ajoudani;Nikos G. Tsagarakis
Ground reaction force information, which includes the location of the center of pressure (COP) and vertical ground reaction force (vGRF), has various applications, such as in the gait assessment of patients post-injury or in the control of robot prostheses and exoskeleton devices. At the beginning of this work, we introduce a newly developed force-sensing device for measuring the COP and vGRF. Then, a model-free calibration method is proposed, leveraging Gaussian process regression (GPR) to extract COP and vGRF from raw sensor data. This approach yields remarkably low normalized root mean squared errors (NRMSEs) of 0.029 and 0.020 for COP in the mediolateral and anteroposterior directions, respectively, and 0.024 for vGRF. However, in general, learning-based calibration methods are sensitive to abnormal readings from sensing elements. To improve the robustness of the measurement, a GPR-based fault detection network is outlined for evaluating the sensing state within the fault in individual sensing elements of the force-sensing device. Moreover, a GPR-based recovery method is proposed to retrieve the sensing device's function under the fault conditions. In validation experiments, the effect of the scale factor of the threshold in the fault detection network is experimentally analyzed. The fault detection network can achieve over 90% success rate with a lower than 5 seconds delay on average in detecting the fault when the scale factor is between 1.68 and 1.90. The engagement of GPR-based recovery models under fault conditions demonstrates a substantial enhancement in COP (up to 85.0% improvement) and vGRF (up to 84.8% improvement) estimation accuracy.
{"title":"On the Calibration, Fault Detection and Recovery of a Force Sensing Device","authors":"Yifang Zhang;Arash Ajoudani;Nikos G. Tsagarakis","doi":"10.1109/LRA.2024.3458807","DOIUrl":"https://doi.org/10.1109/LRA.2024.3458807","url":null,"abstract":"Ground reaction force information, which includes the location of the center of pressure (COP) and vertical ground reaction force (vGRF), has various applications, such as in the gait assessment of patients post-injury or in the control of robot prostheses and exoskeleton devices. At the beginning of this work, we introduce a newly developed force-sensing device for measuring the COP and vGRF. Then, a model-free calibration method is proposed, leveraging Gaussian process regression (GPR) to extract COP and vGRF from raw sensor data. This approach yields remarkably low normalized root mean squared errors (NRMSEs) of 0.029 and 0.020 for COP in the mediolateral and anteroposterior directions, respectively, and 0.024 for vGRF. However, in general, learning-based calibration methods are sensitive to abnormal readings from sensing elements. To improve the robustness of the measurement, a GPR-based fault detection network is outlined for evaluating the sensing state within the fault in individual sensing elements of the force-sensing device. Moreover, a GPR-based recovery method is proposed to retrieve the sensing device's function under the fault conditions. In validation experiments, the effect of the scale factor of the threshold in the fault detection network is experimentally analyzed. The fault detection network can achieve over 90% success rate with a lower than 5 seconds delay on average in detecting the fault when the scale factor is between 1.68 and 1.90. The engagement of GPR-based recovery models under fault conditions demonstrates a substantial enhancement in COP (up to 85.0% improvement) and vGRF (up to 84.8% improvement) estimation accuracy.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10675440","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246504","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-10DOI: 10.1109/LRA.2024.3457383
Lorenzo Montano-Oliván;Julio A. Placed;Luis Montano;María T. Lázaro
Localization in already mapped environments is a critical component in many robotics and automotive applications, where previously acquired information can be exploited along with sensor fusion to provide robust and accurate localization estimates. In this letter, we offer a new perspective on map-based localization by reusing prior topological and metric information. Thus, we reformulate this long-studied problem to go beyond the mere use of metric maps. Our framework seamlessly integrates LiDAR, inertial and GNSS measurements, and cloud-to-map registrations in a sliding window graph fashion, which allows to accommodate the uncertainty of each observation. The modularity of our framework allows it to work with different sensor configurations (e.g., LiDAR resolutions, GNSS denial) and environmental conditions (e.g., mapless regions, large environments). We have conducted several validation experiments, including the deployment in a real-world automotive application, demonstrating the accuracy, efficiency, and versatility of our system in online localization.
{"title":"G-Loc: Tightly-Coupled Graph Localization With Prior Topo-Metric Information","authors":"Lorenzo Montano-Oliván;Julio A. Placed;Luis Montano;María T. Lázaro","doi":"10.1109/LRA.2024.3457383","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457383","url":null,"abstract":"Localization in already mapped environments is a critical component in many robotics and automotive applications, where previously acquired information can be exploited along with sensor fusion to provide robust and accurate localization estimates. In this letter, we offer a new perspective on map-based localization by reusing prior topological and metric information. Thus, we reformulate this long-studied problem to go beyond the mere use of metric maps. Our framework seamlessly integrates LiDAR, inertial and GNSS measurements, and cloud-to-map registrations in a sliding window graph fashion, which allows to accommodate the uncertainty of each observation. The modularity of our framework allows it to work with different sensor configurations (e.g., LiDAR resolutions, GNSS denial) and environmental conditions (e.g., mapless regions, large environments). We have conducted several validation experiments, including the deployment in a real-world automotive application, demonstrating the accuracy, efficiency, and versatility of our system in online localization.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-10DOI: 10.1109/LRA.2024.3457208
Felix Herrmann;Sebastian Zach;Jacopo Banfi;Jan Peters;Georgia Chalvatzaki;Davide Tateo
Estimating collision probabilities between robots and environmental obstacles or other moving agents is crucial to ensure safety during path planning. This is an important building block of modern planning algorithms in many application scenarios such as autonomous driving, where noisy sensors perceive obstacles. While many approaches exist, they either provide too conservative estimates of the collision probabilities or are computationally intensive due to their sampling-based nature. To deal with these issues, we introduce Deep Collision Probability Fields, a neural-based approach for computing collision probabilities of arbitrary objects with arbitrary unimodal uncertainty distributions. Our approach relegates the computationally intensive estimation of collision probabilities via sampling at the training step, allowing for fast neural network inference of the constraints during planning. In extensive experiments, we show that Deep Collision Probability Fields can produce reasonably accurate collision probabilities (up to $10^{-3}$ ) for planning and that our approach can be easily plugged into standard path planning approaches to plan safe paths on 2-D maps containing uncertain static and dynamic obstacles.
{"title":"Safe and Efficient Path Planning Under Uncertainty via Deep Collision Probability Fields","authors":"Felix Herrmann;Sebastian Zach;Jacopo Banfi;Jan Peters;Georgia Chalvatzaki;Davide Tateo","doi":"10.1109/LRA.2024.3457208","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457208","url":null,"abstract":"Estimating collision probabilities between robots and environmental obstacles or other moving agents is crucial to ensure safety during path planning. This is an important building block of modern planning algorithms in many application scenarios such as autonomous driving, where noisy sensors perceive obstacles. While many approaches exist, they either provide too conservative estimates of the collision probabilities or are computationally intensive due to their sampling-based nature. To deal with these issues, we introduce Deep Collision Probability Fields, a neural-based approach for computing collision probabilities of arbitrary objects with arbitrary unimodal uncertainty distributions. Our approach relegates the computationally intensive estimation of collision probabilities via sampling at the training step, allowing for fast neural network inference of the constraints during planning. In extensive experiments, we show that Deep Collision Probability Fields can produce reasonably accurate collision probabilities (up to \u0000<inline-formula><tex-math>$10^{-3}$</tex-math></inline-formula>\u0000) for planning and that our approach can be easily plugged into standard path planning approaches to plan safe paths on 2-D maps containing uncertain static and dynamic obstacles.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142274913","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-10DOI: 10.1109/LRA.2024.3457370
Gyu-Min Oh;Seung-Woo Seo
Precise and real-time localization is crucial for autonomous vehicles. State-of-the-art methods utilize 3D light detection and ranging (LiDAR), inertial measurement unit (IMU), and global positioning system (GPS). However, to meet real-time constraints, these methods often limit the search space to only three degrees of freedom (DoF; $x$ , $y$ , and $heading$ ) and rely on prior maps and IMU for estimating the $roll$ , $pitch$ , and $z$ coordinates. This reliance on maps and sensors can introduce inaccuracies if they contain errors. To achieve precise localization in scenarios where IMU or map errors are present, the $roll$ , $pitch$ , and $z$ coordinates must be estimated. However, incorporating these additional dimensions into the localization process may increase the processing time, rendering it unsuitable for real-time applications. Herein, we propose a precise and robust 6-DoF LiDAR localization algorithm. Instead of directly generating all 6-DoF, the proposed algorithm generates particles based on the $x$ , $y$ , and $heading$ coordinates. Subsequently, it optimizes the estimation of $roll$ , $pitch$ , and $z$ coordinates of each particle while maintaining a fixed number of particles. By expanding the dimensionality in this manner, we mitigate the accuracy degradation that may occur with 3-DoF positioning when dealing with faulty sensors or maps. Experimental results demonstrate that the proposed algorithm achieves satisfactory performance even in scenarios where sensor accuracy is compromised.
{"title":"Fast and Robust 6-DoF LiDAR-Based Localization of an Autonomous Vehicle Against Sensor Inaccuracy","authors":"Gyu-Min Oh;Seung-Woo Seo","doi":"10.1109/LRA.2024.3457370","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457370","url":null,"abstract":"Precise and real-time localization is crucial for autonomous vehicles. State-of-the-art methods utilize 3D light detection and ranging (LiDAR), inertial measurement unit (IMU), and global positioning system (GPS). However, to meet real-time constraints, these methods often limit the search space to only three degrees of freedom (DoF; \u0000<inline-formula><tex-math>$x$</tex-math></inline-formula>\u0000, \u0000<inline-formula><tex-math>$y$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$heading$</tex-math></inline-formula>\u0000) and rely on prior maps and IMU for estimating the \u0000<inline-formula><tex-math>$roll$</tex-math></inline-formula>\u0000, \u0000<inline-formula><tex-math>$pitch$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$z$</tex-math></inline-formula>\u0000 coordinates. This reliance on maps and sensors can introduce inaccuracies if they contain errors. To achieve precise localization in scenarios where IMU or map errors are present, the \u0000<inline-formula><tex-math>$roll$</tex-math></inline-formula>\u0000, \u0000<inline-formula><tex-math>$pitch$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$z$</tex-math></inline-formula>\u0000 coordinates must be estimated. However, incorporating these additional dimensions into the localization process may increase the processing time, rendering it unsuitable for real-time applications. Herein, we propose a precise and robust 6-DoF LiDAR localization algorithm. Instead of directly generating all 6-DoF, the proposed algorithm generates particles based on the \u0000<inline-formula><tex-math>$x$</tex-math></inline-formula>\u0000, \u0000<inline-formula><tex-math>$y$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$heading$</tex-math></inline-formula>\u0000 coordinates. Subsequently, it optimizes the estimation of \u0000<inline-formula><tex-math>$roll$</tex-math></inline-formula>\u0000, \u0000<inline-formula><tex-math>$pitch$</tex-math></inline-formula>\u0000, and \u0000<inline-formula><tex-math>$z$</tex-math></inline-formula>\u0000 coordinates of each particle while maintaining a fixed number of particles. By expanding the dimensionality in this manner, we mitigate the accuracy degradation that may occur with 3-DoF positioning when dealing with faulty sensors or maps. Experimental results demonstrate that the proposed algorithm achieves satisfactory performance even in scenarios where sensor accuracy is compromised.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142235922","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-10DOI: 10.1109/LRA.2024.3457379
Hongxi Wang;Haoxiang Luo;Wei Zhang;Hua Chen
Thanks to recent explosive developments of data-driven learning methodologies, reinforcement learning (RL) emerges as a promising solution to address the legged locomotion problem in robotics. In this letter, we propose CTS, a novel Concurrent Teacher-Student reinforcement learning architecture for legged locomotion over uneven terrains. Different from conventional teacher-student architecture that trains the teacher policy via RL first and then transfers the knowledge to the student policy through supervised learning, our proposed architecture trains teacher and student policy networks concurrently under the reinforcement learning paradigm. To this end, we develop a new training scheme based on a modified proximal policy gradient (PPO) method that exploits data samples collected from the interactions between both the teacher and the student policies with the environment. The effectiveness of the proposed architecture and the new training scheme is demonstrated through substantial quantitative simulation comparisons with the state-of-the-art approaches and extensive indoor and outdoor experiments with quadrupedal and point-foot bipedal robot platforms, showcasing robust and agile locomotion capability. Quantitative simulation comparisons show that our approach reduces the average velocity tracking error by up to 20% compared to the two-stage teacher-student, demonstrating significant superiority in addressing blind locomotion tasks.
{"title":"CTS: Concurrent Teacher-Student Reinforcement Learning for Legged Locomotion","authors":"Hongxi Wang;Haoxiang Luo;Wei Zhang;Hua Chen","doi":"10.1109/LRA.2024.3457379","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457379","url":null,"abstract":"Thanks to recent explosive developments of data-driven learning methodologies, reinforcement learning (RL) emerges as a promising solution to address the legged locomotion problem in robotics. In this letter, we propose CTS, a novel Concurrent Teacher-Student reinforcement learning architecture for legged locomotion over uneven terrains. Different from conventional teacher-student architecture that trains the teacher policy via RL first and then transfers the knowledge to the student policy through supervised learning, our proposed architecture trains teacher and student policy networks concurrently under the reinforcement learning paradigm. To this end, we develop a new training scheme based on a modified proximal policy gradient (PPO) method that exploits data samples collected from the interactions between both the teacher and the student policies with the environment. The effectiveness of the proposed architecture and the new training scheme is demonstrated through substantial quantitative simulation comparisons with the state-of-the-art approaches and extensive indoor and outdoor experiments with quadrupedal and point-foot bipedal robot platforms, showcasing robust and agile locomotion capability. Quantitative simulation comparisons show that our approach reduces the average velocity tracking error by up to 20% compared to the two-stage teacher-student, demonstrating significant superiority in addressing blind locomotion tasks.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This letter presents LiDAR-BIND, a novel sensor fusion framework aimed at enhancing the reliability and safety of autonomous vehicles (AVs) through a shared latent embedding space. With this method, the addition of different modalities, such as sonar and radar, into existing navigation setups becomes possible. These modalities offer robust performance even in challenging scenarios where optical sensors fail. Leveraging a shared latent representation space, LiDAR-BIND enables accurate modality prediction, allowing for the translation of one sensor's observations into another, thereby overcoming the limitations of depending solely on LiDAR for dense point-cloud generation. Through this, the framework facilitates the alignment of multiple sensor modalities without the need for large synchronized datasets across all sensors. We demonstrate its usability in SLAM applications, outperforming traditional LiDAR-based approaches under degraded optical conditions.
本文介绍了一种新型传感器融合框架--LiDAR-BIND，旨在通过共享潜在嵌入空间提高自动驾驶汽车（AV）的可靠性和安全性。有了这种方法，将声纳和雷达等不同模式添加到现有导航设置中就成为可能。即使在光学传感器失效的挑战性场景中，这些模式也能提供强大的性能。利用共享的潜在表示空间，LiDAR-BIND 可实现精确的模态预测，允许将一种传感器的观测结果转换为另一种传感器的观测结果，从而克服了仅依靠激光雷达生成密集点云的局限性。通过这种方法，该框架可促进多种传感器模态的对齐，而无需所有传感器的大型同步数据集。我们展示了该框架在 SLAM 应用中的可用性，在光学条件退化的情况下，其性能优于传统的基于激光雷达的方法。
{"title":"LiDAR-BIND: Multi-Modal Sensor Fusion Through Shared Latent Embeddings","authors":"Niels Balemans;Ali Anwar;Jan Steckel;Siegfried Mercelis","doi":"10.1109/LRA.2024.3457384","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457384","url":null,"abstract":"This letter presents LiDAR-BIND, a novel sensor fusion framework aimed at enhancing the reliability and safety of autonomous vehicles (AVs) through a shared latent embedding space. With this method, the addition of different modalities, such as sonar and radar, into existing navigation setups becomes possible. These modalities offer robust performance even in challenging scenarios where optical sensors fail. Leveraging a shared latent representation space, LiDAR-BIND enables accurate modality prediction, allowing for the translation of one sensor's observations into another, thereby overcoming the limitations of depending solely on LiDAR for dense point-cloud generation. Through this, the framework facilitates the alignment of multiple sensor modalities without the need for large synchronized datasets across all sensors. We demonstrate its usability in SLAM applications, outperforming traditional LiDAR-based approaches under degraded optical conditions.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246489","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-10DOI: 10.1109/LRA.2024.3457385
Louis Wiesmann;Thomas Läbe;Lucas Nunes;Jens Behley;Cyrill Stachniss
Basically all multi-sensor systems must calibrate their sensors to exploit their full potential for state estimation such as mapping and localization. In this letter, we investigate the problem of extrinsic and intrinsic calibration of perception systems. Traditionally, targets in the form of checkerboards or uniquely identifiable tags are used to calibrate those systems. We propose to use a whole calibration environment as a target that supports the intrinsic and extrinsic calibration of different types of sensors. By doing so, we are able to calibrate multiple perception systems with different configurations, sensor types, and sensor modalities. Our approach does not rely on overlaps between sensors which is often otherwise required when using classical targets. The main idea is to relate the measurements for each sensor to a precise model of the calibration environment. For this, we can choose for each sensor a specific method that best suits its calibration. Then, we estimate all intrinsics and extrinsics jointly using least squares adjustment. For the final evaluation of a LiDAR-to-camera calibration of our system, we propose an evaluation method that is independent of the calibration. This allows for quantitative evaluation between different calibration methods. The experiments show that our proposed method is able to provide reliable calibration.
{"title":"Joint Intrinsic and Extrinsic Calibration of Perception Systems Utilizing a Calibration Environment","authors":"Louis Wiesmann;Thomas Läbe;Lucas Nunes;Jens Behley;Cyrill Stachniss","doi":"10.1109/LRA.2024.3457385","DOIUrl":"https://doi.org/10.1109/LRA.2024.3457385","url":null,"abstract":"Basically all multi-sensor systems must calibrate their sensors to exploit their full potential for state estimation such as mapping and localization. In this letter, we investigate the problem of extrinsic and intrinsic calibration of perception systems. Traditionally, targets in the form of checkerboards or uniquely identifiable tags are used to calibrate those systems. We propose to use a whole calibration environment as a target that supports the intrinsic and extrinsic calibration of different types of sensors. By doing so, we are able to calibrate multiple perception systems with different configurations, sensor types, and sensor modalities. Our approach does not rely on overlaps between sensors which is often otherwise required when using classical targets. The main idea is to relate the measurements for each sensor to a precise model of the calibration environment. For this, we can choose for each sensor a specific method that best suits its calibration. Then, we estimate all intrinsics and extrinsics jointly using least squares adjustment. For the final evaluation of a LiDAR-to-camera calibration of our system, we propose an evaluation method that is independent of the calibration. This allows for quantitative evaluation between different calibration methods. The experiments show that our proposed method is able to provide reliable calibration.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142235925","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-09DOI: 10.1109/LRA.2024.3455802
Domen Tabernik;Jon Muhovič;Matej Urbas;Danijel Skočaj
Object grasping is a fundamental challenge in robotics and computer vision, critical for advancing robotic manipulation capabilities. Deformable objects, like fabrics and cloths, pose additional challenges due to their non-rigid nature. In this work, we introduce CeDiRNet-3DoF, a deep-learning model for grasp point detection, with a particular focus on cloth objects. CeDiRNet-3DoF employs center direction regression alongside a localization network, attaining first place in the perception task of ICRA 2023’s Cloth Manipulation Challenge. Recognizing the lack of standardized benchmarks in the literature that hinder effective method comparison, we present the ViCoS Towel Dataset. This extensive benchmark dataset comprises 8,000 real and 12,000 synthetic images, serving as a robust resource for training and evaluating contemporary data-driven deep-learning approaches. Extensive evaluation revealed CeDiRNet-3DoF's robustness in real-world performance, outperforming state-of-the-art methods, including the latest transformer-based models. Our work bridges a crucial gap, offering a robust solution and benchmark for cloth grasping in computer vision and robotics.
{"title":"Center Direction Network for Grasping Point Localization on Cloths","authors":"Domen Tabernik;Jon Muhovič;Matej Urbas;Danijel Skočaj","doi":"10.1109/LRA.2024.3455802","DOIUrl":"https://doi.org/10.1109/LRA.2024.3455802","url":null,"abstract":"Object grasping is a fundamental challenge in robotics and computer vision, critical for advancing robotic manipulation capabilities. Deformable objects, like fabrics and cloths, pose additional challenges due to their non-rigid nature. In this work, we introduce CeDiRNet-3DoF, a deep-learning model for grasp point detection, with a particular focus on cloth objects. CeDiRNet-3DoF employs center direction regression alongside a localization network, attaining first place in the perception task of ICRA 2023’s Cloth Manipulation Challenge. Recognizing the lack of standardized benchmarks in the literature that hinder effective method comparison, we present the ViCoS Towel Dataset. This extensive benchmark dataset comprises 8,000 real and 12,000 synthetic images, serving as a robust resource for training and evaluating contemporary data-driven deep-learning approaches. Extensive evaluation revealed CeDiRNet-3DoF's robustness in real-world performance, outperforming state-of-the-art methods, including the latest transformer-based models. Our work bridges a crucial gap, offering a robust solution and benchmark for cloth grasping in computer vision and robotics.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142230874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-09-09DOI: 10.1109/LRA.2024.3456509
Simone Ferrari;Luca Di Giammarino;Leonardo Brizi;Giorgio Grisetti
LiDAR odometry is the task of estimating the ego-motion of the sensor from sequential laser scans. This problem has been addressed by the community for more than two decades, and many effective solutions are available nowadays. Most of these systems implicitly rely on assumptions about the operating environment, the sensor used, and motion pattern. When these assumptions are violated, several well-known systems tend to perform poorly. This letter presents a LiDAR odometry system that can overcome these limitations and operate well under different operating conditions while achieving performance comparable with domain-specific methods. Our algorithm follows the well-known ICP paradigm that leverages a PCA-based kd-tree implementation that is used to extract structural information about the clouds being registered and to compute the minimization metric for the alignment. The drift is bound by managing the local map based on the estimated uncertainty of the tracked pose.
{"title":"MAD-ICP: It is All About Matching Data – Robust and Informed LiDAR Odometry","authors":"Simone Ferrari;Luca Di Giammarino;Leonardo Brizi;Giorgio Grisetti","doi":"10.1109/LRA.2024.3456509","DOIUrl":"https://doi.org/10.1109/LRA.2024.3456509","url":null,"abstract":"LiDAR odometry is the task of estimating the ego-motion of the sensor from sequential laser scans. This problem has been addressed by the community for more than two decades, and many effective solutions are available nowadays. Most of these systems implicitly rely on assumptions about the operating environment, the sensor used, and motion pattern. When these assumptions are violated, several well-known systems tend to perform poorly. This letter presents a LiDAR odometry system that can overcome these limitations and operate well under different operating conditions while achieving performance comparable with domain-specific methods. Our algorithm follows the well-known ICP paradigm that leverages a PCA-based kd-tree implementation that is used to extract structural information about the clouds being registered and to compute the minimization metric for the alignment. The drift is bound by managing the local map based on the estimated uncertainty of the tracked pose.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":null,"pages":null},"PeriodicalIF":4.6,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142246506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：