Pub Date : 2025-12-27DOI: 10.1016/j.robot.2025.105321
Rafał Staszak, Bartlomiej Kulecki, Marek Kraft, Dominik Belter
Robots face challenges in perceiving new scenes, particularly when registering objects from a single perspective, resulting in incomplete shape information about objects. Partial object models negatively influence the performance of grasping methods. To address this, robots can scan the scene from various perspectives or employ methods to directly fill in unknown regions. This research reexamines scene reconstruction typically formulated in 3D space, proposing a novel formulation in 2D image space for robots with RGB-D cameras. We introduce a method that generates a depth image from a virtual camera pose located on the opposite position of the reconstructed object. The article demonstrates that the convolutional neural network can be trained for accurate depth image generation and subsequent 3D scene reconstruction from a single viewpoint. We show that the proposed approach is computationally efficient and accurate when compared to methods that operate directly in 3D space. Furthermore, we illustrate the application of this model in enhancing grasping method success rates.
{"title":"MirrorNet: Hallucinating 2.5D depth images for efficient 3D scene reconstruction","authors":"Rafał Staszak, Bartlomiej Kulecki, Marek Kraft, Dominik Belter","doi":"10.1016/j.robot.2025.105321","DOIUrl":"10.1016/j.robot.2025.105321","url":null,"abstract":"<div><div>Robots face challenges in perceiving new scenes, particularly when registering objects from a single perspective, resulting in incomplete shape information about objects. Partial object models negatively influence the performance of grasping methods. To address this, robots can scan the scene from various perspectives or employ methods to directly fill in unknown regions. This research reexamines scene reconstruction typically formulated in 3D space, proposing a novel formulation in 2D image space for robots with RGB-D cameras. We introduce a method that generates a depth image from a virtual camera pose located on the opposite position of the reconstructed object. The article demonstrates that the convolutional neural network can be trained for accurate depth image generation and subsequent 3D scene reconstruction from a single viewpoint. We show that the proposed approach is computationally efficient and accurate when compared to methods that operate directly in 3D space. Furthermore, we illustrate the application of this model in enhancing grasping method success rates.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105321"},"PeriodicalIF":5.2,"publicationDate":"2025-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885479","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 : 2025-12-27DOI: 10.1016/j.robot.2025.105322
Mengtang Li , Shen Zhao , Shuai Wang , Fanmao Liu
In conventional minimally invasive surgery, an assistant manually steers the endoscope based on the surgeon’s verbal commands, but fatigue and tremor can degrade field-of-view (FOV) stability and efficiency. Robotic endoscopes address this limitation through automated FOV adjustment via image-based visual servoing, ensuring smooth and stable visualization. However, most robotic implementations mount rigid straight-rod endoscopes on external serial arms, limiting dexterity and complicating remote-center-of-motion (RCM) control. Moreover, many automated FOV methods track surgical-tool tips without representing the surgeon’s intention. This work therefore presents a compact 6-DOF parallel endoscopic mechanism that improves flexibility and dexterity while simplifying RCM constraint satisfaction, together with an eye-gaze-assisted multi-tool tracking controller that dynamically weights tools according to surgeon attention. Simulations and experiments across diverse scenarios demonstrate FOV stabilization within 2 s, mean image-space tracking error 20 pixels, eye hand error 3°, and at least a 30% reduction in unnecessary FOV adjustments. Supplementary video is available.
{"title":"Enhanced flexibility and dexterity in robotic endoscopy via a 6-DOF parallel mechanism and eye-gaze-assisted field-of-view control","authors":"Mengtang Li , Shen Zhao , Shuai Wang , Fanmao Liu","doi":"10.1016/j.robot.2025.105322","DOIUrl":"10.1016/j.robot.2025.105322","url":null,"abstract":"<div><div>In conventional minimally invasive surgery, an assistant manually steers the endoscope based on the surgeon’s verbal commands, but fatigue and tremor can degrade field-of-view (FOV) stability and efficiency. Robotic endoscopes address this limitation through automated FOV adjustment via image-based visual servoing, ensuring smooth and stable visualization. However, most robotic implementations mount rigid straight-rod endoscopes on external serial arms, limiting dexterity and complicating remote-center-of-motion (RCM) control. Moreover, many automated FOV methods track surgical-tool tips without representing the surgeon’s intention. This work therefore presents a compact 6-DOF parallel endoscopic mechanism that improves flexibility and dexterity while simplifying RCM constraint satisfaction, together with an eye-gaze-assisted multi-tool tracking controller that dynamically weights tools according to surgeon attention. Simulations and experiments across diverse scenarios demonstrate FOV stabilization within 2 s, mean image-space tracking error <span><math><mo><</mo></math></span> 20 pixels, eye hand error <span><math><mo><</mo></math></span> 3°, and at least a 30% reduction in unnecessary FOV adjustments. Supplementary video is available.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105322"},"PeriodicalIF":5.2,"publicationDate":"2025-12-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885484","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 : 2025-12-26DOI: 10.1016/j.robot.2025.105320
Jingyi Liu , Hengyu Li , Hang Liu , Shaorong Xie , Jun Luo
Image inpainting is a critical technique for recovering missing information caused by camera soiling on mobile robots. However, most existing learning-based methods still struggle to handle damaged images with complex semantic environments and diverse hole patterns, primarily because of the insufficient acquisition and inadequate fusion of scene-consistent prior cues for damaged images. To address this limitation, we propose a novel reference-guided image inpainting network () for mobile robots equipped with binocular cameras, which employs adjacent camera images as inpainting guidance and fuses its prior information via progressive feature interaction to reconstruct damaged regions. Specifically, a back-projection-based feature interaction module (FIM) is proposed to align the features of the reference and damaged images, thereby capturing the contextual information of the reference image for inpainting. Additionally, a content reconstruction module (CRM) based on residual learning and channel attention is presented to selectively aggregate interactive features for reconstructing missing details. Building upon these two modules, we further devise a progressive feature interaction and reconstruction module (PFIRM) that organizes multiple FIM-CRM pairs into a stepwise structure, enabling the progressive fusion of multiscale contextual information derived from both the damaged and reference images. Moreover, a feature refinement module (FRM) is developed to interact with low-level fine-grained features and refine the reconstructed details. Extensive evaluations conducted on the public ETHZ dataset and our self-built MII dataset demonstrate that outperforms other state-of-the-art approaches and produces high-quality inpainting results on real soiled data.
{"title":"Reference-guided image inpainting via progressive feature interaction and reconstruction for mobile robots with binocular cameras","authors":"Jingyi Liu , Hengyu Li , Hang Liu , Shaorong Xie , Jun Luo","doi":"10.1016/j.robot.2025.105320","DOIUrl":"10.1016/j.robot.2025.105320","url":null,"abstract":"<div><div>Image inpainting is a critical technique for recovering missing information caused by camera soiling on mobile robots. However, most existing learning-based methods still struggle to handle damaged images with complex semantic environments and diverse hole patterns, primarily because of the insufficient acquisition and inadequate fusion of scene-consistent prior cues for damaged images. To address this limitation, we propose a novel reference-guided image inpainting network (<span><math><mrow><msup><mrow><mi>RGI</mi></mrow><mrow><mn>2</mn></mrow></msup><mi>N</mi></mrow></math></span>) for mobile robots equipped with binocular cameras, which employs adjacent camera images as inpainting guidance and fuses its prior information via progressive feature interaction to reconstruct damaged regions. Specifically, a back-projection-based feature interaction module (FIM) is proposed to align the features of the reference and damaged images, thereby capturing the contextual information of the reference image for inpainting. Additionally, a content reconstruction module (CRM) based on residual learning and channel attention is presented to selectively aggregate interactive features for reconstructing missing details. Building upon these two modules, we further devise a progressive feature interaction and reconstruction module (PFIRM) that organizes multiple FIM-CRM pairs into a stepwise structure, enabling the progressive fusion of multiscale contextual information derived from both the damaged and reference images. Moreover, a feature refinement module (FRM) is developed to interact with low-level fine-grained features and refine the reconstructed details. Extensive evaluations conducted on the public ETHZ dataset and our self-built MII dataset demonstrate that <span><math><mrow><msup><mrow><mi>RGI</mi></mrow><mrow><mn>2</mn></mrow></msup><mi>N</mi></mrow></math></span> outperforms other state-of-the-art approaches and produces high-quality inpainting results on real soiled data.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105320"},"PeriodicalIF":5.2,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145842778","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}
Task and Motion Planning (TAMP) is essential for efficient Human–Robot Collaboration (HRC) in industrial settings, yet existing approaches struggle to handle human interventions and dynamic environments. This paper presents a Human Adaptive Task and Motion Planning (HAD-TAMP) framework that seamlessly integrates human pose and actions into the planning process to quickly adapt to human requests or deviations from the process plan. The framework consists of three key modules: a task planning module, which generates and updates task sequences based on real-time human input, a motion planning module composed of a set of motion planners specialized for different phases of the collaboration (e.g., collaborative transportation of materials), and a context reasoner module which coordinates the overall process based on the sensory information available. A key contribution is using a receding horizon strategy, enabling real-time adaptation to human inputs and environmental changes. The approach is validated in a real industrial HRC scenario through two applications: gesture-based human–robot interaction and close human–robot collaboration in carbon fiber draping. Experimental results demonstrate the framework’s effectiveness in ensuring adaptability to multiple human requests and efficiency: the re-planning time is 4 times and 5 times faster than the generation of a new plan.
{"title":"HAD-TAMP: Human adaptive task and motion planning for human–robot collaboration in industrial scenario","authors":"Alberto Gottardi , Matteo Terreran , Enrico Pagello , Emanuele Menegatti","doi":"10.1016/j.robot.2025.105318","DOIUrl":"10.1016/j.robot.2025.105318","url":null,"abstract":"<div><div>Task and Motion Planning (TAMP) is essential for efficient Human–Robot Collaboration (HRC) in industrial settings, yet existing approaches struggle to handle human interventions and dynamic environments. This paper presents a Human Adaptive Task and Motion Planning (HAD-TAMP) framework that seamlessly integrates human pose and actions into the planning process to quickly adapt to human requests or deviations from the process plan. The framework consists of three key modules: a task planning module, which generates and updates task sequences based on real-time human input, a motion planning module composed of a set of motion planners specialized for different phases of the collaboration (e.g., collaborative transportation of materials), and a context reasoner module which coordinates the overall process based on the sensory information available. A key contribution is using a receding horizon strategy, enabling real-time adaptation to human inputs and environmental changes. The approach is validated in a real industrial HRC scenario through two applications: gesture-based human–robot interaction and close human–robot collaboration in carbon fiber draping. Experimental results demonstrate the framework’s effectiveness in ensuring adaptability to multiple human requests and efficiency: the re-planning time is 4 times and 5 times faster than the generation of a new plan.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105318"},"PeriodicalIF":5.2,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885483","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 : 2025-12-24DOI: 10.1016/j.robot.2025.105316
Pan Li , Yifei Chen , Xinghua Lin , Chongcong Ye , Junxia Zhang , Delei Fang , Cunman Liang
This paper systematically reviews the research advancements, core technical characteristics, and application challenges of magnetic continuum robots. As a key direction in the field of soft robotics, magnetic continuum robots leverage the non-contact manipulation characteristics of magnetic field actuation to demonstrate unique advantages in narrow-space operations (such as intravascular guidewire navigation), high-precision manipulation (such as minimally invasive surgery), and dynamic environmental adaptation. Their application scenarios have expanded from single intravascular interventions to complex task execution across the entire body. The paper highlights that the development of various magnetic actuators (such as gradient magnetic field generators and rotating magnetic field devices) has accelerated improvements in the robots’ motion flexibility and task adaptability. However, significant technical bottlenecks remain: insufficient environmental adaptability of control algorithms leading to trajectory deviations, instability caused by the complexity of system dynamics modeling, contradictions between the spatial uniformity and penetration depth of the actuation magnetic field, and the challenge of balancing biocompatibility, mechanical durability, and magnetic response efficiency in flexible polymer materials—all of which limit their clinical application. Finally, from the perspective of technological integration and breakthroughs, this paper discusses future development directions: deep integration of artificial intelligence and control algorithms, application of biocompatible materials and 3D printing technologies, optimization of magnetic actuation platforms, enhancement of multimodal collaborative operation capabilities, and expansion into interdisciplinary fields such as environmental monitoring and elderly care services. This review provides a systematic research framework and conceptual references for the technological evolution and practical application of magnetic continuum robots.
{"title":"A survey on magnetically driven continuum robots for biomedical application","authors":"Pan Li , Yifei Chen , Xinghua Lin , Chongcong Ye , Junxia Zhang , Delei Fang , Cunman Liang","doi":"10.1016/j.robot.2025.105316","DOIUrl":"10.1016/j.robot.2025.105316","url":null,"abstract":"<div><div>This paper systematically reviews the research advancements, core technical characteristics, and application challenges of magnetic continuum robots. As a key direction in the field of soft robotics, magnetic continuum robots leverage the non-contact manipulation characteristics of magnetic field actuation to demonstrate unique advantages in narrow-space operations (such as intravascular guidewire navigation), high-precision manipulation (such as minimally invasive surgery), and dynamic environmental adaptation. Their application scenarios have expanded from single intravascular interventions to complex task execution across the entire body. The paper highlights that the development of various magnetic actuators (such as gradient magnetic field generators and rotating magnetic field devices) has accelerated improvements in the robots’ motion flexibility and task adaptability. However, significant technical bottlenecks remain: insufficient environmental adaptability of control algorithms leading to trajectory deviations, instability caused by the complexity of system dynamics modeling, contradictions between the spatial uniformity and penetration depth of the actuation magnetic field, and the challenge of balancing biocompatibility, mechanical durability, and magnetic response efficiency in flexible polymer materials—all of which limit their clinical application. Finally, from the perspective of technological integration and breakthroughs, this paper discusses future development directions: deep integration of artificial intelligence and control algorithms, application of biocompatible materials and 3D printing technologies, optimization of magnetic actuation platforms, enhancement of multimodal collaborative operation capabilities, and expansion into interdisciplinary fields such as environmental monitoring and elderly care services. This review provides a systematic research framework and conceptual references for the technological evolution and practical application of magnetic continuum robots.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105316"},"PeriodicalIF":5.2,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885480","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 : 2025-12-23DOI: 10.1016/j.robot.2025.105311
Abdalkarim Mohtasib , Heriberto Cuayáhuitl
Neural-based reinforcement learning is a promising approach for teaching robots new behaviours. But one of its main limitations is the need for carefully hand-coded reward signals by an expert. It is thus crucial to automate the reward learning process so that new skills can be taught to robots by their users. This article proposes an approach for enabling robots to learn reward signals for sequential tasks from visual observations, eliminating the need for expert-designed reward signals. It involves dividing the sequential task into smaller sub-tasks using a novel auto-labelling technique to generate rewards for demonstration data. A novel image classifier is proposed to estimate the visual rewards for each task accurately. The effectiveness of the proposed approach is demonstrated in generating informative reward signals through comprehensive evaluations on three challenging sequential tasks: block stacking, door opening, and nuts assembly. By using the learnt reward signals to train reinforcement learning agents from demonstration, we are able to induce policies that outperform those trained with sparse oracle rewards. Since our approach consistently outperformed several baselines including DDPG, TD3, SAC, DAPG, GAIL, and AWAC, it represents an advancement in the application of model-free reinforcement learning to sequential robotic tasks.
{"title":"Robot policy learning from demonstrations and visual rewards for sequential manipulation tasks","authors":"Abdalkarim Mohtasib , Heriberto Cuayáhuitl","doi":"10.1016/j.robot.2025.105311","DOIUrl":"10.1016/j.robot.2025.105311","url":null,"abstract":"<div><div>Neural-based reinforcement learning is a promising approach for teaching robots new behaviours. But one of its main limitations is the need for carefully hand-coded reward signals by an expert. It is thus crucial to automate the reward learning process so that new skills can be taught to robots by their users. This article proposes an approach for enabling robots to learn reward signals for sequential tasks from visual observations, eliminating the need for expert-designed reward signals. It involves dividing the sequential task into smaller sub-tasks using a novel auto-labelling technique to generate rewards for demonstration data. A novel image classifier is proposed to estimate the visual rewards for each task accurately. The effectiveness of the proposed approach is demonstrated in generating informative reward signals through comprehensive evaluations on three challenging sequential tasks: block stacking, door opening, and nuts assembly. By using the learnt reward signals to train reinforcement learning agents from demonstration, we are able to induce policies that outperform those trained with sparse oracle rewards. Since our approach consistently outperformed several baselines including DDPG, TD3, SAC, DAPG, GAIL, and AWAC, it represents an advancement in the application of model-free reinforcement learning to sequential robotic tasks.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105311"},"PeriodicalIF":5.2,"publicationDate":"2025-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145842779","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 : 2025-12-23DOI: 10.1016/j.robot.2025.105314
Dan Chen , Kaiwen Luo , Zichen Wang , Lintao Fan , Changqing Wang
In order to solve the problems of strong expansion randomness, slow convergence speed, and poor environmental adaptability in the optimal Rapidly-exploring Random Tree (RRT*) algorithm, this paper proposes an adaptive bidirectional RRT*(AB-RRT*) algorithm. The algorithm first initializes the target bias probability and expansion step size in random sampling using an evaluation function based on factors such as map size, number of obstacles, and starting point distance. It adaptively adjusts the target bias probability and expansion step size based on the number of collision detections, reducing the randomness of the sampling process and effectively reducing the number of redundant points generated. Secondly, collision zone shielding and extended node guidance strategies were proposed to guide random trees quickly through narrow and complex environments, improving the convergence speed of the algorithm. Finally, redundant points are removed and smoothed from the initial path to obtain a better global path. We conducted comparative experiments between the proposed algorithm and RRT*, Q-RRT*, GB-RRT, RRT-Connect, RRT*-Connect, GBB-RRT*, and Informed-Bi-RRT* in four different complexity scenarios. The results showed that compared with the best performing Informed-Bi-RRT* algorithm, the AB-RRT* algorithm shortened the path generation time by 17.2% to 71.54% in different scenarios, reduced the path length by 0.44% to 2.25%, and achieved a planning success rate of 100%. This proves the superiority of AB-RRT* algorithm in planning efficiency and path quality, as well as its excellent adaptability and robustness in different environments.
针对最优快速探索随机树（RRT*）算法存在的扩展随机性强、收敛速度慢、环境适应性差等问题，本文提出了一种自适应双向RRT*(AB-RRT*)算法。该算法首先利用基于地图大小、障碍物数量和起点距离等因素的评价函数初始化随机抽样中的目标偏差概率和扩展步长。该算法根据碰撞检测次数自适应调整目标偏置概率和展开步长，降低了采样过程的随机性，有效减少了生成的冗余点数量。其次，提出碰撞区屏蔽和扩展节点引导策略，引导随机树快速通过狭窄复杂的环境，提高了算法的收敛速度；最后，从初始路径中去除冗余点并进行平滑处理，得到更好的全局路径。在四种不同的复杂度场景下，将本文算法与RRT*、Q-RRT*、GB-RRT、RRT-Connect、RRT*-Connect、GBB-RRT*、Informed-Bi-RRT*进行了对比实验。结果表明，与表现最佳的inform - bi - rrt *算法相比，AB-RRT*算法在不同场景下的路径生成时间缩短了17.2%至71.54%，路径长度缩短了0.44%至2.25%，规划成功率为100%。这证明了AB-RRT*算法在规划效率和路径质量上的优越性，以及在不同环境下良好的适应性和鲁棒性。
{"title":"An adaptive bidirectional optimal rapidly exploring random tree algorithm with dynamic adjustment and extended guidance strategy","authors":"Dan Chen , Kaiwen Luo , Zichen Wang , Lintao Fan , Changqing Wang","doi":"10.1016/j.robot.2025.105314","DOIUrl":"10.1016/j.robot.2025.105314","url":null,"abstract":"<div><div>In order to solve the problems of strong expansion randomness, slow convergence speed, and poor environmental adaptability in the optimal Rapidly-exploring Random Tree (RRT*) algorithm, this paper proposes an adaptive bidirectional RRT*(AB-RRT*) algorithm. The algorithm first initializes the target bias probability and expansion step size in random sampling using an evaluation function based on factors such as map size, number of obstacles, and starting point distance. It adaptively adjusts the target bias probability and expansion step size based on the number of collision detections, reducing the randomness of the sampling process and effectively reducing the number of redundant points generated. Secondly, collision zone shielding and extended node guidance strategies were proposed to guide random trees quickly through narrow and complex environments, improving the convergence speed of the algorithm. Finally, redundant points are removed and smoothed from the initial path to obtain a better global path. We conducted comparative experiments between the proposed algorithm and RRT*, Q-RRT*, GB-RRT, RRT-Connect, RRT*-Connect, GBB-RRT*, and Informed-Bi-RRT* in four different complexity scenarios. The results showed that compared with the best performing Informed-Bi-RRT* algorithm, the AB-RRT* algorithm shortened the path generation time by 17.2% to 71.54% in different scenarios, reduced the path length by 0.44% to 2.25%, and achieved a planning success rate of 100%. This proves the superiority of AB-RRT* algorithm in planning efficiency and path quality, as well as its excellent adaptability and robustness in different environments.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105314"},"PeriodicalIF":5.2,"publicationDate":"2025-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885485","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 : 2025-12-22DOI: 10.1016/j.robot.2025.105313
Nisha Kumari, Kevin Lee, Chathu Ranaweera
Drone swarms operate as decentralized systems where multiple autonomous nodes coordinate their actions through inter-drone communication. A network is a collection of interconnected nodes that communicate to share resources, with its topology representing the physical or logical arrangement of these nodes. For drone swarms, network topology plays a key role in enabling coordinated actions through effective communication links. Understanding the behavior of drone swarms requires analyzing their network topology, as it provides valuable insights into the links and nodes that define their communication patterns. The research in this paper presents a computer vision-based approach to extract and analyze the network topology of such swarms, focusing on the logical communication links rather than physical formations. Using 3D coordinates obtained via stereo vision, the method identifies communication patterns corresponding to star, ring and mesh topologies. The experimental results demonstrate that the proposed method can accurately distinguish between different communication patterns within the swarm, allowing for effective mapping of the network structure. This analysis provides practical insights into how swarm coordination emerges from communication topology and offers a foundation for optimizing swarm behavior in real-world applications.
{"title":"Visually extracting the network topology of drone swarms","authors":"Nisha Kumari, Kevin Lee, Chathu Ranaweera","doi":"10.1016/j.robot.2025.105313","DOIUrl":"10.1016/j.robot.2025.105313","url":null,"abstract":"<div><div>Drone swarms operate as decentralized systems where multiple autonomous nodes coordinate their actions through inter-drone communication. A network is a collection of interconnected nodes that communicate to share resources, with its topology representing the physical or logical arrangement of these nodes. For drone swarms, network topology plays a key role in enabling coordinated actions through effective communication links. Understanding the behavior of drone swarms requires analyzing their network topology, as it provides valuable insights into the links and nodes that define their communication patterns. The research in this paper presents a computer vision-based approach to extract and analyze the network topology of such swarms, focusing on the logical communication links rather than physical formations. Using 3D coordinates obtained via stereo vision, the method identifies communication patterns corresponding to star, ring and mesh topologies. The experimental results demonstrate that the proposed method can accurately distinguish between different communication patterns within the swarm, allowing for effective mapping of the network structure. This analysis provides practical insights into how swarm coordination emerges from communication topology and offers a foundation for optimizing swarm behavior in real-world applications.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105313"},"PeriodicalIF":5.2,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145808533","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 : 2025-12-22DOI: 10.1016/j.robot.2025.105315
Roun Lee , Seonghun Hong , Sukmin Yoon
Over the past few decades, light detection and ranging (LiDAR) sensors have been extensively employed for pose estimation in simultaneous localization and mapping (SLAM). In more recent years, the use of solid-state LiDAR sensors with no rotating mechanisms and a limited field-of-view for SLAM has attracted research attention because of their cost effectiveness and durability. However, it is highly challenging to successfully perform place recognition, which is one of the most important components of SLAM, via limited field-of-view measurements. Failure in place recognition can severely degrade the resulting estimation performance of SLAM algorithms. Considering a terrestrial SLAM framework for quadruped robots with limited field-of-view LiDAR sensors, this study proposes a terrain-based place recognition algorithm that reconstructs and compares detected feature terrains, using a set of foot contact information for quadruped robots. The validity and practical feasibility of the proposed approach are demonstrated through experimental results using a quadruped robot system with a limited field-of-view LiDAR sensor.
{"title":"Terrain-based place recognition for LiDAR SLAM of quadruped robots with limited field-of-view measurements","authors":"Roun Lee , Seonghun Hong , Sukmin Yoon","doi":"10.1016/j.robot.2025.105315","DOIUrl":"10.1016/j.robot.2025.105315","url":null,"abstract":"<div><div>Over the past few decades, light detection and ranging (LiDAR) sensors have been extensively employed for pose estimation in simultaneous localization and mapping (SLAM). In more recent years, the use of solid-state LiDAR sensors with no rotating mechanisms and a limited field-of-view for SLAM has attracted research attention because of their cost effectiveness and durability. However, it is highly challenging to successfully perform place recognition, which is one of the most important components of SLAM, via limited field-of-view measurements. Failure in place recognition can severely degrade the resulting estimation performance of SLAM algorithms. Considering a terrestrial SLAM framework for quadruped robots with limited field-of-view LiDAR sensors, this study proposes a terrain-based place recognition algorithm that reconstructs and compares detected feature terrains, using a set of foot contact information for quadruped robots. The validity and practical feasibility of the proposed approach are demonstrated through experimental results using a quadruped robot system with a limited field-of-view LiDAR sensor.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105315"},"PeriodicalIF":5.2,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939734","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 : 2025-12-21DOI: 10.1016/j.robot.2025.105312
Si-Hao Zuo , Yan-Jiang Zhao , Yong-De Zhang , Le Ye
Frame transformation is the basis in the robotic system, and the topology structural linkage between the frames is the foundation of the frame transformation. However, the frame transformation process for a complex robotic system is rather challenging and complicated because there may exist a lot of frames to be defined and linked in different technical scenarios, which are inexplicit and individual. This severely hinders the exchanges between the engineers and hampers the development of robotic technologies as well. This paper proposes a novel framework to explicitly describe the linkages between the frames, and automatically realize a complete frame transformation for the robotic system. The framework involves three layers: the semantic description layer, the frame topology relationship layer, and the mathematical calculation layer, and it achieves a unity of the three layers. We define an element module to semantically describe each object with a fixed frame, design a relative position transformation chain (RPTC) to explicitly describe the topology structural linkages between the frames. Then, an expansionary and evolutionary strategy is proposed to obtain a final result of the RPTC for the robotic system by the expansions and evolutions of the short RPTCs. Finally, a software platform is developed for the realization of the framework, and a case of a classical medical robotic system is performed as an example. And the results of the expansion and the evolution degrees prove the validity of the proposed method.
{"title":"An expansion and evolution framework of frame transformation for complex robotic systems","authors":"Si-Hao Zuo , Yan-Jiang Zhao , Yong-De Zhang , Le Ye","doi":"10.1016/j.robot.2025.105312","DOIUrl":"10.1016/j.robot.2025.105312","url":null,"abstract":"<div><div>Frame transformation is the basis in the robotic system, and the topology structural linkage between the frames is the foundation of the frame transformation. However, the frame transformation process for a complex robotic system is rather challenging and complicated because there may exist a lot of frames to be defined and linked in different technical scenarios, which are inexplicit and individual. This severely hinders the exchanges between the engineers and hampers the development of robotic technologies as well. This paper proposes a novel framework to explicitly describe the linkages between the frames, and automatically realize a complete frame transformation for the robotic system. The framework involves three layers: the semantic description layer, the frame topology relationship layer, and the mathematical calculation layer, and it achieves a unity of the three layers. We define an element module to semantically describe each object with a fixed frame, design a relative position transformation chain (RPTC) to explicitly describe the topology structural linkages between the frames. Then, an expansionary and evolutionary strategy is proposed to obtain a final result of the RPTC for the robotic system by the expansions and evolutions of the short RPTCs. Finally, a software platform is developed for the realization of the framework, and a case of a classical medical robotic system is performed as an example. And the results of the expansion and the evolution degrees prove the validity of the proposed method.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105312"},"PeriodicalIF":5.2,"publicationDate":"2025-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940040","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号