Robotics and Autonomous Systems最新文献

{"title":"Adaptive neural network fault-tolerant control of tendon actuated continuum robots under actuator failures and external disturbances","authors":"Hongyun Liu,&nbsp;Weidong Liu,&nbsp;Jinbo Zhong","doi":"10.1016/j.robot.2025.105310","DOIUrl":"10.1016/j.robot.2025.105310","url":null,"abstract":"<div><div>This paper proposes an adaptive neural network backstepping sliding mode fault-tolerant control (ANN-BSMFTC) strategy to enhance the control robustness of tendon-actuated continuum robots (TACR) under actuator failures and sudden external disturbances. Initially, a precise dynamic model of TACR is established based on Cosserat rod theory. Subsequently, a backstepping control framework is employed to derive the system’s virtual control quantities, with sliding mode control (SMC) embedded to strengthen robustness against parameter perturbations and external disturbances. Furthermore, a radial basis function neural network (RBFNN) is incorporated to online approximate unknown components in the control law through its global approximation capability. The asymptotic convergence of the closed-loop system is rigorously proven under concurrent partial actuator failure and external step disturbance conditions using Lyapunov stability theory. Comparative simulation experiments with sliding mode fault-tolerant control (SMFTC), backstepping fault-tolerant control (BFTC), and backstepping sliding mode fault-tolerant control (BSMFTC) validate the effectiveness. Results demonstrate that the proposed ANN-BSMFTC strategy achieves optimal spatial trajectory tracking performance, maintaining the average normalized root mean square error (NRMSE) of position tracking below 5% under extreme conditions involving actuator faults and sudden disturbances.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105310"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885372","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}

{"title":"Curriculum-guided deep reinforcement learning with fuzzy rewards for autonomous push-grasp manipulation in cluttered environments","authors":"Chih-Yung Huang ,&nbsp;Adhan Efendi ,&nbsp;Ying-Chun Wang","doi":"10.1016/j.robot.2025.105325","DOIUrl":"10.1016/j.robot.2025.105325","url":null,"abstract":"<div><div>Robotic manipulation in cluttered environments requires robust coordination of pushing and grasping to overcome occlusions, constrained grasp geometries, and uncertain object interactions. This study presents a curriculum-guided deep reinforcement learning framework that jointly redesigns the training distribution, state abstraction, and reward structure for autonomous push–grasp manipulation. A depth-aware grasp potential module constructs a geometric affordance map that prioritizes feasible top-layer grasp opportunities, guiding the agent toward collision-free rearrangement behaviors. A fuzzy logic–based reward mechanism integrates changes in graspable area and grasp Q-values into a continuous shaping signal, addressing sparse feedback and stabilizing learning. A stage-wise curriculum with proportion-controlled difficulty progression gradually increases clutter density and object difficulty, enabling progressive acquisition of coordinated push–grasp skills. Extensive evaluations across randomized clutter, structured challenge scenarios, and real-world experiments on previously unseen and semi-transparent objects show that the proposed framework consistently outperforms VPG-based and grasp-quality baselines in grasp success and action efficiency. These results demonstrate the effectiveness of coupling curriculum design, depth-aware grasp prioritization, and fuzzy reward shaping for robust manipulation in complex, cluttered settings.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105325"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037689","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}

{"title":"AdaRRT: A novel adaptive path planning algorithm for mobile robots in complex terrain environments","authors":"Yuqing Chen ,&nbsp;Haotong Yan ,&nbsp;Shiyu Li ,&nbsp;Shunqing Yang ,&nbsp;Huosheng Hu","doi":"10.1016/j.robot.2026.105354","DOIUrl":"10.1016/j.robot.2026.105354","url":null,"abstract":"<div><div>Path planning for mobile robots in complex environments has attracted increasing attention. However, the classical RRT and its variants still suffer from low sampling efficiency, unsafe paths close to obstacles, and susceptibility to local optima. Recently proposed methods, such as Adapted-RRT, QS-RRT*, and PQ-RRT*, partially address these issues but rely on complex operators and parameter tuning, leading to high computational costs and poor real-time suitability. To address these limitations, we propose an Adaptive RRT (AdaRRT). Its qualitative advantages include adaptive exploration via dynamic sampling adjustment, improved safety through obstacle-avoiding growth, enhanced robustness with local-optima escape, and smoother paths by tangent-based optimization. Quantitatively, simulations show that AdaRRT reduces search time by up to 91.14%, shortens path length by 20-33%, and decreases node count by over 70%. Real-world experiments further validate its efficiency and safety. Overall, AdaRRT outperforms existing methods in efficiency, safety, and robustness, offering a practical solution for autonomous navigation in complex environments.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105354"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037684","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}

{"title":"RPV-Map: A novel Reflection-Planes Voxel Map for glass and mirror reconstruction on SLAM System","authors":"Hao-Tien Yeh ,&nbsp;Jun-Jie Hu ,&nbsp;Fu-Hao Chang ,&nbsp;Kuan-Ting Lin ,&nbsp;Li-Chen Fu","doi":"10.1016/j.robot.2025.105308","DOIUrl":"10.1016/j.robot.2025.105308","url":null,"abstract":"<div><div>This paper presents a novel reflection-plane voxel map approach that effectively detects and reconstructs reflective objects during SLAM by integrating geometric and visual information from 3D LiDAR and camera. The method first estimates rough reflection planes from the voxel map and extracts 3D boundary and surface points through intensity-based point cloud classification. Using 2D Alpha Shape boundary extraction and multi-view reflection mask validation, we achieve reliable geometric feature extraction. Our proposed plane optimization method considers weighted distributions of different point types and employs boundary refinement to improve reconstruction accuracy. Experimental results demonstrate that our approach outperforms existing methods in terms of plane accuracy, reconstruction completeness, and point cloud classification while achieving real-time performance.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105308"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145808532","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}

{"title":"Robust predefined-time target surrounding control for multiple underactuated unmanned surface vessels","authors":"Xuecheng Zhou ,&nbsp;Xuehong Tian ,&nbsp;Qingqun Mai ,&nbsp;Jing Zhang ,&nbsp;Haitao Liu","doi":"10.1016/j.robot.2025.105309","DOIUrl":"10.1016/j.robot.2025.105309","url":null,"abstract":"<div><div>A robust predefined-time targeted surrounding fault-tolerant controller is proposed in this paper for multiple underactuated unmanned surface vessels (USVs) to rotate uniformly around a target vessel with all errors converging in a predefined time. First, a predefined-time distributed state observer (PTDSO) is designed to observe the states of the target USV, which are passed to other USVs through the communication topology. Second, the case of input saturation is considered, a saturation function is constructed to constrain the control signals, and a predefined-time auxiliary dynamic system is constructed to compensate for the controller. Similarly, the velocities of the USVs are indirectly constrained by constraining the virtual control signals. Third, to enhance the robustness of the controller, the disturbances to which the USV is subjected are divided into two parts: general disturbances and strong disturbances. A predefined-time adaptive fuzzy neural network (FNN) is designed to approximate general environmental disturbances and model uncertainties and actuator faults, and a predefined-time <span><math><msub><mi>H</mi><mi>∞</mi></msub></math></span> algorithm is designed to suppress strong disturbances. Finally, the stability analysis proves that all the signals within the closed-loop system can converge within a predefined time. Numerical simulations demonstrate the effectiveness of the proposed algorithm.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105309"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885482","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}

{"title":"Reference-guided image inpainting via progressive feature interaction and reconstruction for mobile robots with binocular cameras","authors":"Jingyi Liu ,&nbsp;Hengyu Li ,&nbsp;Hang Liu ,&nbsp;Shaorong Xie ,&nbsp;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":"2026-04-01","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}

{"title":"Hybrid attention-guided RRT*: Learning spatial sampling priors for accelerated path planning","authors":"Asmaa Loulou ,&nbsp;Mustafa Unel","doi":"10.1016/j.robot.2026.105338","DOIUrl":"10.1016/j.robot.2026.105338","url":null,"abstract":"<div><div>Sampling-based planners such as RRT* are widely used for motion planning in high-dimensional and complex environments. However, their reliance on uniform sampling often leads to slow convergence and inefficiency, especially in scenarios with narrow passages or long-range dependencies. To address this, we propose HAGRRT*, a Hybrid Attention-Guided RRT* algorithm that learns to generate spatially informed sampling priors. Our method introduces a new neural architecture that fuses multi-scale convolutional features with a lightweight cross-attention mechanism, explicitly conditioned on the start and goal positions. These features are decoded via a DPT-inspired module to produce 2D probability maps that guide the sampling process. Additionally, we propose an obstacle-aware loss function that penalizes disconnected and infeasible predictions which further encourages the network to focus on traversable, goal-directed regions. Extensive experiments on both structured (maze) and unstructured (forest) environments show that HAGRRT* achieves significantly faster convergence and improved path quality compared to both classical RRT* and recent deep-learning guided variants. Our method consistently requires fewer iterations and samples and is able to generalize across varying dataset types. On structured scenarios, our method achieves an average reduction of <strong>39.6%</strong> in the number of samples and an average of <strong>24.4%</strong> reduction in planning time compared to recent deep learning methods. On unstructured forest maps, our method reduces the number of samples by <strong>71.5%</strong>, and planning time by <strong>81.7%</strong> compared to recent deep learning methods, and improves the success rate from <strong>67%</strong> to <strong>93%</strong>. These results highlight the robustness, efficiency, and generalization ability of our approach across a wide range of planning environments.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105338"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978328","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}

{"title":"Adaptive artificial potential field method for small autonomous vehicles","authors":"Kemal Ihsan Kilic,&nbsp;Aurélien Desoeuvres,&nbsp;Casper Bak Pedersen,&nbsp;Alex Elkjær Vasegaard,&nbsp;Peter Nielsen","doi":"10.1016/j.robot.2026.105364","DOIUrl":"10.1016/j.robot.2026.105364","url":null,"abstract":"<div><div>Artificial Potential Field (APF) is one of the path planning and obstacle avoidance methods used for its simplicity and effectiveness. The goal’s attractive force and the obstacles’ repulsive forces are modeled and considered to act upon the vehicle or robot. However, in practice, the classical APF faces challenges like local minima. We propose to study several enhancements and their combinations in order to show how they can create even more efficient algorithms to overcome these challenges. Those enhancements include tangential force, inertia-inspired force, and a local minima detection (l.m.d.) and reaction scheme by adding virtual obstacles and dynamically changing coefficients. All of them keep the methods as lightweight as possible while improving the classical APF. The tangential force provides smoother paths and avoids local minima cases. The dynamic change of APF’s parameters, coupled with the addition of virtual obstacles when detecting local minima, provides an efficient way to escape them. The inertia-inspired force can be used to smooth the trajectory when only obstacles in front of the vehicle are taken into account. We defined performance metrics to assess the path completion, path quality, and processing time to compare the proposed enhancements with the base case of classical APF. We benchmarked the proposed methods in different environments for a holonomic robot and a simplified bicycle. The proposed adaptive APF with inertial force extension completed 87.5% of the tests while the classical APF completed only 43.8% of them. On the other side, tangential versions of APF reduce the path length deviation by 8% and the curvature by 20% in simple cases. The code is available on: <span><span>https://github.com/Glawal/APFproject/tree/paper1</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105364"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037686","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}

{"title":"Global navigation-local operation: Pose optimization and multi-sensory guidance strategy for autonomous mobile manipulators","authors":"Dianfan Zhang ,&nbsp;Yujie Zhu ,&nbsp;Shuhong Cheng ,&nbsp;Chao Zhang ,&nbsp;Zedai Wang ,&nbsp;Shijun Zhang","doi":"10.1016/j.robot.2025.105307","DOIUrl":"10.1016/j.robot.2025.105307","url":null,"abstract":"<div><div>This paper presents a novel control method for integrating global navigation and local operation in an autonomous mobile operating system (AMOS). A pose optimization method based on null-space projection is proposed to enable seamless switching between global navigation and local operation. Additionally, a control strategy that combines both visual perception and force feedback for effective guidance during local operations is introduced. Firstly, utilizing projected gradient, the task space is partitioned into primary and null spaces, and secondary task functions are designed to optimize and constrain the pose while preserving the primary task. Secondly, a 2-stage guidance strategy for shaft hole insertion, incorporating visual perception and force control techniques, is presented. This strategy involves visual guidance for point cloud registration and fine adjustment using force sensing. To address transient and steady-state errors in the system, a finite-time prescribed performance super-twisting algorithm (PPSTA) controller is proposed and its stability is proven using Lyapunov theory. Simulation and experimental results validate the effectiveness of the proposed method in achieving both navigation and manipulation tasks for mobile manipulators with high performance and robustness.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105307"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145808531","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}

{"title":"Vision-guided gripping process with minimizing folding for flexible fabric materials by integrating a sequential optimization algorithm and FEM analysis","authors":"Minh Khang Ngo ,&nbsp;Chanhee Won ,&nbsp;HyunKyo Lim ,&nbsp;Dae Young Lim ,&nbsp;Than Trong Khanh Dat ,&nbsp;Jonghun Yoon","doi":"10.1016/j.robot.2026.105349","DOIUrl":"10.1016/j.robot.2026.105349","url":null,"abstract":"<div><div>Automatic fabric handling in garment manufacturing presents significant challenges due to the soft, deformable, and highly variable nature of textile materials. This paper proposes an integrated robotic fabric gripping system capable of reliably identifying and manipulating fabric parts with minimal folding. The system comprises an industrial robotic arm, four needle grippers mounted on an adjustable jig mechanism, and a high-precision 3D vision camera for real-time fabric detection. A key contribution of this work is a sequential optimization model that determines four optimal gripping points for each fabric pattern, based on material characterization, deformation criteria, and folding definitions derived from experiments and finite element simulations. These points are mapped relative to CAD data and stored for retrieval. A vision-based matching algorithm then aligns real-time image inputs with CAD templates to localize the fabric piece and recover the precomputed optimal gripping points, which are transmitted to the robot for autonomous execution. Quantitative evaluations demonstrate that the proposed approach significantly reduces fabric folding and enhances the reliability of robotic garment handling, representing a substantial step toward fully automated garment production.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"198 ","pages":"Article 105349"},"PeriodicalIF":5.2,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978333","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}