Pub Date : 2024-12-26DOI: 10.1109/LRA.2024.3522844
Fangyan Zheng;Shuai Xin;Xinghui Han;Lin Hua
The parallel kinematic mechanism (PKM) is typically equipped with multiple actuators to realize the precise and arbitrary spatial motion, but resulting in complex mechanical and control systems and high cost. Actually, in many specific fields, such as heavy load metal forming process, the motion of PKM is only needed to be specific and the motion precision requirement is not extremely high. This paper proposes a new approximate mechanism synthesis method for PKM with single actuator (PKMSA), which not only can simplify the complexity of PKM and reduce the cost, but also can realize the specific motion with permissible error. Firstly, the design criteria of the consistency between the motion pattern of the PKMSA and required motion DoF is determined to avoid the occurrence of kinematic redundancy error. Then a general kinematic model for PKMSA is derived based on the screw theory, and the general constraints are obtained for PKMSA to realize the specific motion. On this basis, a 3-RSS/S PKMSA configured with a single input and triple output actuator layout realized by a gear set is proposed. Finally, a heavy load multi-DoF forming machine (load of 200 kN) with PKMSA is developed and, with that, the multi-DoF forming experiment of a typical metal component is conducted with forming load of about 180 kN. The geometric deviation of formed component is in the range of −40∼55 μm (it can completely meet the forming accuracy requirement), validating the feasibility of the proposed approximate mechanism synthesis method for PKMSA.
{"title":"A Novel Parallel Kinematic Mechanism With Single Actuator for Multi-DoF Forming Machine","authors":"Fangyan Zheng;Shuai Xin;Xinghui Han;Lin Hua","doi":"10.1109/LRA.2024.3522844","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522844","url":null,"abstract":"The parallel kinematic mechanism (PKM) is typically equipped with multiple actuators to realize the precise and arbitrary spatial motion, but resulting in complex mechanical and control systems and high cost. Actually, in many specific fields, such as heavy load metal forming process, the motion of PKM is only needed to be specific and the motion precision requirement is not extremely high. This paper proposes a new approximate mechanism synthesis method for PKM with single actuator (PKMSA), which not only can simplify the complexity of PKM and reduce the cost, but also can realize the specific motion with permissible error. Firstly, the design criteria of the consistency between the motion pattern of the PKMSA and required motion DoF is determined to avoid the occurrence of kinematic redundancy error. Then a general kinematic model for PKMSA is derived based on the screw theory, and the general constraints are obtained for PKMSA to realize the specific motion. On this basis, a 3-RSS/S PKMSA configured with a single input and triple output actuator layout realized by a gear set is proposed. Finally, a heavy load multi-DoF forming machine (load of 200 kN) with PKMSA is developed and, with that, the multi-DoF forming experiment of a typical metal component is conducted with forming load of about 180 kN. The geometric deviation of formed component is in the range of −40∼55 μm (it can completely meet the forming accuracy requirement), validating the feasibility of the proposed approximate mechanism synthesis method for PKMSA.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1832-1839"},"PeriodicalIF":4.6,"publicationDate":"2024-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142976001","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 : 2024-12-26DOI: 10.1109/LRA.2024.3522836
Emil Stubbe Kolvig-Raun;Jakob Hviid;Mikkel Baun Kjærgaard;Ralph Brorsen;Peter Jacob
In our experience, the task of optimizing robot longevity and efficiency is challenging due to the limited understanding and awareness developers' have about how their code influences a robot's expected lifespan. Unfortunately, acquiring the necessary information for computations is a complex task, and the data needed for these calculations remains unattainable until after runtime. In software engineering, traditional Static Code Analysis (SCA) techniques are applied to address such challenges. Although effective in identifying software anomalies and inefficiencies without execution, current SCA techniques do not adequately address the unique requirements of Cyber-Physical Systems (CPSs) in robotics. In this study, we propose a novel Machine Learning (ML) approach to assess robot program lines, considering the balance between speed and lifespan. Our solution, trained on data from 1325 operational collaborative robots (cobots) from the Universal Robots (UR) e-Series, classifies program lines concerning the expected lifespan of the robot, considering program line arguments, expected resource usage, and asserted joint stress. The model achieves a worst-case accuracy of 90.43% through 10-fold cross-validation with a 50% data split. We also present a selection of programming lines illustrating various robot program cases and an example of longevity improvement. Finally, we publish a dataset containing 56405 unique program line executions, aiming to enhance the sustainability and efficiency of robotic systems and support future research.
{"title":"Balancing Cobot Productivity and Longevity Through Pre-Runtime Developer Feedback","authors":"Emil Stubbe Kolvig-Raun;Jakob Hviid;Mikkel Baun Kjærgaard;Ralph Brorsen;Peter Jacob","doi":"10.1109/LRA.2024.3522836","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522836","url":null,"abstract":"In our experience, the task of optimizing robot longevity and efficiency is challenging due to the limited understanding and awareness developers' have about how their code influences a robot's expected lifespan. Unfortunately, acquiring the necessary information for computations is a complex task, and the data needed for these calculations remains unattainable until after runtime. In software engineering, traditional Static Code Analysis (SCA) techniques are applied to address such challenges. Although effective in identifying software anomalies and inefficiencies without execution, current SCA techniques do not adequately address the unique requirements of Cyber-Physical Systems (CPSs) in robotics. In this study, we propose a novel Machine Learning (ML) approach to assess robot program lines, considering the balance between speed and lifespan. Our solution, trained on data from 1325 operational collaborative robots (cobots) from the Universal Robots (UR) e-Series, classifies program lines concerning the expected lifespan of the robot, considering program line arguments, expected resource usage, and asserted joint stress. The model achieves a worst-case accuracy of 90.43% through 10-fold cross-validation with a 50% data split. We also present a selection of programming lines illustrating various robot program cases and an example of longevity improvement. Finally, we publish a dataset containing 56405 unique program line executions, aiming to enhance the sustainability and efficiency of robotic systems and support future research.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1617-1624"},"PeriodicalIF":4.6,"publicationDate":"2024-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142940753","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 : 2024-12-25DOI: 10.1109/LRA.2024.3522840
Jun Hou;Shiyu Xing;Yunkai Ma;Fengshui Jing;Min Tan
Robotic pose servoing aims to move the robot end-effector to the target pose. Closed-loop servo systems can tolerate a small TCF (tool control frame) calibration error and accurately reach the target pose through multiple pose measurements and pose adjustments. However, the maximum allowable TCF calibration error remains an open question. This paper demonstrates that the necessary condition for robotic pose servoing is a TCF calibration error angle of less than 60 degrees, with no limit on the translational component of the TCF calibration error. Next, an improved pose servoing method is proposed to address the conflict between the large TCF error and the limited robot workspace. This method introduces a scaling factor to limit the adjustment range within the robot workspace, ensuring greater robustness. Finally, robot-assisted cabin docking is selected as an experimental validation case. Simulation and physical experiments validate the maximum allowable TCF calibration error. Comparative experiments confirm the robustness of the improved pose servoing method, achieving cabin docking despite significant TCF calibration errors.
{"title":"Maximum Allowable TCF Calibration Error for Robotic Pose Servoing","authors":"Jun Hou;Shiyu Xing;Yunkai Ma;Fengshui Jing;Min Tan","doi":"10.1109/LRA.2024.3522840","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522840","url":null,"abstract":"Robotic pose servoing aims to move the robot end-effector to the target pose. Closed-loop servo systems can tolerate a small TCF (tool control frame) calibration error and accurately reach the target pose through multiple pose measurements and pose adjustments. However, the maximum allowable TCF calibration error remains an open question. This paper demonstrates that the necessary condition for robotic pose servoing is a TCF calibration error angle of less than 60 degrees, with no limit on the translational component of the TCF calibration error. Next, an improved pose servoing method is proposed to address the conflict between the large TCF error and the limited robot workspace. This method introduces a scaling factor to limit the adjustment range within the robot workspace, ensuring greater robustness. Finally, robot-assisted cabin docking is selected as an experimental validation case. Simulation and physical experiments validate the maximum allowable TCF calibration error. Comparative experiments confirm the robustness of the improved pose servoing method, achieving cabin docking despite significant TCF calibration errors.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1744-1751"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142976120","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 : 2024-12-25DOI: 10.1109/LRA.2024.3522779
Christine T. Chang;Maria P. Stull;Breanne Crockett;Emily Jensen;Clare Lohrmann;Mitchell Hebert;Bradley Hayes
Frictionless and understandable tasking is essential for leveraging human-autonomy teaming in commercial, military, and public safety applications. Existing technology for facilitating human teaming with uncrewed aerial vehicles (UAVs), utilizing planners or trajectory optimizers that incorporate human input, introduces a usability and operator capability gap by not explicitly effecting user upskilling by promoting system understanding or predictability. Supplementing annotated waypoints with natural language guidance affords an opportunity for both. In this work we investigate one-shot versus iterative input, introducing a testbed system based on government and industry UAV planning tools that affords inputs in the form of both natural language text and drawn annotations on a terrain map. The testbed uses an LLM-based subsystem to map user inputs into additional terms for the trajectory optimization objective function. We demonstrate through a human subjects study that prompting a human teammate to iteratively add latent knowledge to a trajectory optimization aids the user in learning how the system functions, elicits more desirable robot behaviors, and ultimately achieves better task outcomes.
{"title":"Iteratively Adding Latent Human Knowledge Within Trajectory Optimization Specifications Improves Learning and Task Outcomes","authors":"Christine T. Chang;Maria P. Stull;Breanne Crockett;Emily Jensen;Clare Lohrmann;Mitchell Hebert;Bradley Hayes","doi":"10.1109/LRA.2024.3522779","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522779","url":null,"abstract":"Frictionless and understandable tasking is essential for leveraging human-autonomy teaming in commercial, military, and public safety applications. Existing technology for facilitating human teaming with uncrewed aerial vehicles (UAVs), utilizing planners or trajectory optimizers that incorporate human input, introduces a usability and operator capability gap by not explicitly effecting user upskilling by promoting system understanding or predictability. Supplementing annotated waypoints with natural language guidance affords an opportunity for both. In this work we investigate one-shot versus iterative input, introducing a testbed system based on government and industry UAV planning tools that affords inputs in the form of both natural language text and drawn annotations on a terrain map. The testbed uses an LLM-based subsystem to map user inputs into additional terms for the trajectory optimization objective function. We demonstrate through a human subjects study that prompting a human teammate to iteratively add latent knowledge to a trajectory optimization aids the user in learning how the system functions, elicits more desirable robot behaviors, and ultimately achieves better task outcomes.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1537-1544"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142938316","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 : 2024-12-25DOI: 10.1109/LRA.2024.3522773
Junze Yang;Qiuxuan Wu;Shenao Li;Yuejun Ye;Cenfeng Luo
Biped wheel-legged robot is an important configuration of ground mobile robot. In the existing research, in order to deploy the model on an embedded system with limited computing power, the balance control of the robot is usually decoupled from the body posture and steering, which reduces the control coordination of the robot. In order to solve the above problems, firstly, the novel Full-State Dynamics Model(FSDM) is introduced, and the model is linearized by Taylor expansion and solving the limit of multivariate function. Secondly, a novel Forward Kinematics(FK) solution method based on trajectory equation is proposed for Virtual Model Control(VMC). Compared with the general FK solution method, it can further significantly improve the calculation speed of VMC on the embedded platform. Furthermore, the Linear Quadratic Regulator(LQR) controller is optimized, and the weight matrix value can be automatically adjusted according to the error of the state variable. At the same time, simulation results show that the motion performance of the robot can be improved by actively adjusting the posture. Therefore, an adaptive LQR controller, a steering compensator and a gravity compensator are designed. Simulation and physical experimental results verify the effectiveness of the proposed model, controller and control strategy.
{"title":"Integrated Modeling and Control Optimization of Biped Wheel-Legged Robot","authors":"Junze Yang;Qiuxuan Wu;Shenao Li;Yuejun Ye;Cenfeng Luo","doi":"10.1109/LRA.2024.3522773","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522773","url":null,"abstract":"Biped wheel-legged robot is an important configuration of ground mobile robot. In the existing research, in order to deploy the model on an embedded system with limited computing power, the balance control of the robot is usually decoupled from the body posture and steering, which reduces the control coordination of the robot. In order to solve the above problems, firstly, the novel Full-State Dynamics Model(FSDM) is introduced, and the model is linearized by Taylor expansion and solving the limit of multivariate function. Secondly, a novel Forward Kinematics(FK) solution method based on trajectory equation is proposed for Virtual Model Control(VMC). Compared with the general FK solution method, it can further significantly improve the calculation speed of VMC on the embedded platform. Furthermore, the Linear Quadratic Regulator(LQR) controller is optimized, and the weight matrix value can be automatically adjusted according to the error of the state variable. At the same time, simulation results show that the motion performance of the robot can be improved by actively adjusting the posture. Therefore, an adaptive LQR controller, a steering compensator and a gravity compensator are designed. Simulation and physical experimental results verify the effectiveness of the proposed model, controller and control strategy.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1465-1472"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142938304","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}
Although the LiDAR SLAM technique has been already widely deployed on various robots, it may still suffers from degeneracy caused by inadequate constraints in scenes with sparse geometric features. If the degeneracy is not detected and properly processed, the accuracy of localization and mapping will significantly decrease. In this letter, we propose the P2d-DO method, which consists of a point-to-distribution degeneracy detection algorithm and a point cloud-weighted degeneracy optimization algorithm, to relieve the negative impact of degeneracy. The degeneracy detection algorithm outputs factors that characterize the degeneracy state by observing changes in the distribution probabilities within a local region. Factors reflecting the confidence of the point clouds are then fed to the degeneracy optimization algorithm, enabling the system to prioritize reliable point clouds by assigning larger weights during the matching process. Comprehensive experiments validate the effectiveness of our method, demonstrating significant improvements in both degeneracy detection and pose estimation in terms of accuracy and robustness.
{"title":"P2d-DO: Degeneracy Optimization for LiDAR SLAM With Point-to-Distribution Detection Factors","authors":"Weinan Chen;Sehua Ji;Xubin Lin;Zhi-Xin Yang;Wenzheng Chi;Yisheng Guan;Haifei Zhu;Hong Zhang","doi":"10.1109/LRA.2024.3522839","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522839","url":null,"abstract":"Although the LiDAR SLAM technique has been already widely deployed on various robots, it may still suffers from degeneracy caused by inadequate constraints in scenes with sparse geometric features. If the degeneracy is not detected and properly processed, the accuracy of localization and mapping will significantly decrease. In this letter, we propose the P2d-DO method, which consists of a point-to-distribution degeneracy detection algorithm and a point cloud-weighted degeneracy optimization algorithm, to relieve the negative impact of degeneracy. The degeneracy detection algorithm outputs factors that characterize the degeneracy state by observing changes in the distribution probabilities within a local region. Factors reflecting the confidence of the point clouds are then fed to the degeneracy optimization algorithm, enabling the system to prioritize reliable point clouds by assigning larger weights during the matching process. Comprehensive experiments validate the effectiveness of our method, demonstrating significant improvements in both degeneracy detection and pose estimation in terms of accuracy and robustness.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1489-1496"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142938308","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 : 2024-12-25DOI: 10.1109/LRA.2024.3521679
Zhen Deng;Weiwei Liu;Guotao Li;Jianwei Zhang
In this letter, a constrained visual predictive control strategy (C-VPC) is developed for a robotic flexible endoscope to precisely track target features in narrow environments while adhering to visibility and joint limit constraints. The visibility constraint, crucial for keeping the target feature within the camera's field of view, is explicitly designed using zeroing control barrier functions to uphold the forward invariance of a visible set. To automate the robotic endoscope, kinematic modeling for image-based visual servoing is conducted, resulting in a state-space model that facilitates the prediction of the future evolution of the endoscopic state. The C-VPC method calculates the optimal control input by optimizing the model-based predictions of the future state under visibility and joint limit constraints. Both simulation and experimental results demonstrate the effectiveness of the proposed method in achieving autonomous target tracking and addressing the visibility constraint simultaneously. The proposed method achieved a reduction of 12.3% in Mean Absolute Error (MAE) and 56.0% in variance (VA) compared to classic IBVS.
{"title":"Constrained Visual Predictive Control of a Robotic Flexible Endoscope With Visibility and Joint Limits Constraints","authors":"Zhen Deng;Weiwei Liu;Guotao Li;Jianwei Zhang","doi":"10.1109/LRA.2024.3521679","DOIUrl":"https://doi.org/10.1109/LRA.2024.3521679","url":null,"abstract":"In this letter, a constrained visual predictive control strategy (C-VPC) is developed for a robotic flexible endoscope to precisely track target features in narrow environments while adhering to visibility and joint limit constraints. The visibility constraint, crucial for keeping the target feature within the camera's field of view, is explicitly designed using zeroing control barrier functions to uphold the forward invariance of a visible set. To automate the robotic endoscope, kinematic modeling for image-based visual servoing is conducted, resulting in a state-space model that facilitates the prediction of the future evolution of the endoscopic state. The C-VPC method calculates the optimal control input by optimizing the model-based predictions of the future state under visibility and joint limit constraints. Both simulation and experimental results demonstrate the effectiveness of the proposed method in achieving autonomous target tracking and addressing the visibility constraint simultaneously. The proposed method achieved a reduction of 12.3% in Mean Absolute Error (MAE) and 56.0% in variance (VA) compared to classic IBVS.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1513-1520"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142938307","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}
Offline reinforcement learning strives to enable agents to effectively utilize pre-collected offline datasets for learning. Such an offline setup tremendously mitigates the problems of online reinforcement learning algorithms in real-world applications, particularly in scenarios where interactions are constrained or exploration is costly. The learned strategy, on the other hand, has a distributional bias with respect to the behavioral strategy, which consequently leads to the problem of extrapolation error for out-of-distribution actions. To mitigate this problem, in this paper, we adopt a hierarchical offline reinforcement learning framework that extracts recurrent and spatio-temporally extended primitive skills from offline data before using them for downstream task learning. Besides, we introduce an autodecoder conditional diffusion model to characterize low-level strategy decoding. Such a deep learning generative model enables the reduction of action primitives for the strategy space, which is then used to learn high-level task strategy-guided primitives via the offline learning algorithm IQL. Experimental results and ablation studies on D4RL benchmark tasks (Antmaze, Adroit and Kitchen) demonstrate that our approach achieves SOTA performance in most tasks.
{"title":"DASP: Hierarchical Offline Reinforcement Learning via Diffusion Autodecoder and Skill Primitive","authors":"Sicheng Liu;Yunchuan Zhang;Wenbai Chen;Peiliang Wu","doi":"10.1109/LRA.2024.3522842","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522842","url":null,"abstract":"Offline reinforcement learning strives to enable agents to effectively utilize pre-collected offline datasets for learning. Such an offline setup tremendously mitigates the problems of online reinforcement learning algorithms in real-world applications, particularly in scenarios where interactions are constrained or exploration is costly. The learned strategy, on the other hand, has a distributional bias with respect to the behavioral strategy, which consequently leads to the problem of extrapolation error for out-of-distribution actions. To mitigate this problem, in this paper, we adopt a hierarchical offline reinforcement learning framework that extracts recurrent and spatio-temporally extended primitive skills from offline data before using them for downstream task learning. Besides, we introduce an autodecoder conditional diffusion model to characterize low-level strategy decoding. Such a deep learning generative model enables the reduction of action primitives for the strategy space, which is then used to learn high-level task strategy-guided primitives via the offline learning algorithm IQL. Experimental results and ablation studies on D4RL benchmark tasks (Antmaze, Adroit and Kitchen) demonstrate that our approach achieves SOTA performance in most tasks.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1649-1655"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142940706","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 : 2024-12-25DOI: 10.1109/LRA.2024.3522778
Hang Yu;ChristopheDe Wagter;Guido C. H. E de Croon
Autonomous flight in unknown, cluttered environments is still a major challenge in robotics. Existing obstacle avoidance algorithms typically adopt a fixed flight velocity, overlooking the crucial balance between safety and agility. We propose a reinforcement learning algorithm to learn an adaptive flight speed policy tailored to varying environment complexities, enhancing obstacle avoidance safety. A downside of learning-based obstacle avoidance algorithms is that the lack of a mapping module can lead to the drone getting stuck in complex scenarios. To address this, we introduce a novel training setup for the latent space that retains memory of previous depth map observations. The latent space is explicitly trained to predict both past and current depth maps. Our findings confirm that varying speed leads to a superior balance of success rate and agility in cluttered environments. Additionally, our memory-augmented latent representation outperforms the latent representation commonly used in reinforcement learning. Furthermore, an extensive comparison of our method with the existing state-of-the-art approaches Agile-autonomy and Ego-planner shows the superior performance of our approach, especially in highly cluttered environments. Finally, after minimal fine-tuning, we successfully deployed our network on a real drone for enhanced obstacle avoidance.
{"title":"MAVRL: Learn to Fly in Cluttered Environments With Varying Speed","authors":"Hang Yu;ChristopheDe Wagter;Guido C. H. E de Croon","doi":"10.1109/LRA.2024.3522778","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522778","url":null,"abstract":"Autonomous flight in unknown, cluttered environments is still a major challenge in robotics. Existing obstacle avoidance algorithms typically adopt a fixed flight velocity, overlooking the crucial balance between safety and agility. We propose a reinforcement learning algorithm to learn an adaptive flight speed policy tailored to varying environment complexities, enhancing obstacle avoidance safety. A downside of learning-based obstacle avoidance algorithms is that the lack of a mapping module can lead to the drone getting stuck in complex scenarios. To address this, we introduce a novel training setup for the latent space that retains memory of previous depth map observations. The latent space is explicitly trained to predict both past and current depth maps. Our findings confirm that varying speed leads to a superior balance of success rate and agility in cluttered environments. Additionally, our memory-augmented latent representation outperforms the latent representation commonly used in reinforcement learning. Furthermore, an extensive comparison of our method with the existing state-of-the-art approaches Agile-autonomy and Ego-planner shows the superior performance of our approach, especially in highly cluttered environments. Finally, after minimal fine-tuning, we successfully deployed our network on a real drone for enhanced obstacle avoidance.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1441-1448"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142918349","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 : 2024-12-25DOI: 10.1109/LRA.2024.3522848
Jinning Li;Jiachen Li;Sangjae Bae;David Isele
Deep learning-based trajectory prediction models for autonomous driving often struggle with generalization to out-of-distribution (OOD) scenarios, sometimes performing worse than simple rule-based models. To address this limitation, we propose a novel framework, Adaptive Prediction Ensemble (APE), which integrates deep learning and rule-based prediction experts. A learned routing function, trained concurrently with the deep learning model, dynamically selects the most reliable prediction based on the input scenario. Our experiments on large-scale datasets, including Waymo Open Motion Dataset (WOMD) and Argoverse, demonstrate improvement in zero-shot generalization across datasets. We show that our method outperforms individual prediction models and other variants, particularly in long-horizon prediction and scenarios with a high proportion of OOD data. This work highlights the potential of hybrid approaches for robust and generalizable motion prediction in autonomous driving.
{"title":"Adaptive Prediction Ensemble: Improving Out-of-Distribution Generalization of Motion Forecasting","authors":"Jinning Li;Jiachen Li;Sangjae Bae;David Isele","doi":"10.1109/LRA.2024.3522848","DOIUrl":"https://doi.org/10.1109/LRA.2024.3522848","url":null,"abstract":"Deep learning-based trajectory prediction models for autonomous driving often struggle with generalization to out-of-distribution (OOD) scenarios, sometimes performing worse than simple rule-based models. To address this limitation, we propose a novel framework, Adaptive Prediction Ensemble (APE), which integrates deep learning and rule-based prediction experts. A learned routing function, trained concurrently with the deep learning model, dynamically selects the most reliable prediction based on the input scenario. Our experiments on large-scale datasets, including Waymo Open Motion Dataset (WOMD) and Argoverse, demonstrate improvement in zero-shot generalization across datasets. We show that our method outperforms individual prediction models and other variants, particularly in long-horizon prediction and scenarios with a high proportion of OOD data. This work highlights the potential of hybrid approaches for robust and generalizable motion prediction in autonomous driving.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 2","pages":"1553-1560"},"PeriodicalIF":4.6,"publicationDate":"2024-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142938314","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：