{"title":"CBTMP: Optimizing Multi-Agent Path Finding in Heterogeneous Cooperative Environments","authors":"Jianqi Gao;Yanjie Li;Yongjin Mu;Qi Liu;Haoyao Chen;Yunjiang Lou","doi":"10.1109/LRA.2025.3557672","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557672","url":null,"abstract":"This letter introduces the Conflict-Based Three-agent Meeting with Pickup (CBTMP), a near-optimal algorithm tailored for cooperative multi-agent path finding in heterogeneous environments, specifically to boost the operational efficiency of intelligent warehouses. CBTMP is a two-level algorithm. The high-level policy identifies the meeting positions for heterogeneous agents by reformulating the cooperative multi-agent path finding problem as a multi-group, three-agent meeting with pickup problem. Using the meeting positions and predefined task positions, the low-level policy utilizes the proposed conflict-based search with time-step alignment algorithm to plan conflict-free paths for all heterogeneous agents. Extensive evaluations on six two-dimensional grid benchmark maps reveal that CBTMP not only significantly bolsters solution success rates but also attains near-optimal sum-of-costs and makespan values. To confirm its real-world applicability, we also validate CBTMP through experiments with physical Turtlebot3 robots.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5010-5017"},"PeriodicalIF":4.6,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856241","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}
Thomas Pritchard;Saifullah Ijaz;Ronald Clark;Basaran Bahadir Kocer
{"title":"ForestVO: Enhancing Visual Odometry in Forest Environments Through ForestGlue","authors":"Thomas Pritchard;Saifullah Ijaz;Ronald Clark;Basaran Bahadir Kocer","doi":"10.1109/LRA.2025.3557738","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557738","url":null,"abstract":"Recent advancements in visual odometry systems have improved autonomous navigation, yet challenges persist in complex environments like forests, where dense foliage, variable lighting, and repetitive textures compromise the accuracy of feature correspondences. To address these challenges, we introduce ForestGlue. ForestGlue enhances the SuperPoint feature detector through four configurations – grayscale, RGB, RGB-D, and stereo-vision inputs – optimised for various sensing modalities. For feature matching, we employ LightGlue or SuperGlue, both of which have been retrained using synthetic forest data. ForestGlue achieves comparable pose estimation accuracy to baseline LightGlue and SuperGlue models, yet require only 512 keypoints, just 25% of the 2048 keypoints used by baseline models, to achieve an LO-RANSAC AUC score of 0.745 at a 10° threshold. With a 1/4 of the keypoints required, ForestGlue has the potential to reduce computational overhead whilst being effective in dynamic forest environments, making it a promising candidate for real-time deployment on resource-constrained platforms such as drones or mobile robotic platforms. By combining ForestGlue with a novel transformer based pose estimation model, we propose ForestVO, which estimates relative camera poses using the 2D pixel coordinates of matched features between frames. On challenging TartanAir forest sequences, ForestVO achieves an average relative pose error (RPE) of 1.09 m and kitti_score of 2.33%, outperforming direct-based methods such as DSO in dynamic scenes by 40%, while maintaining competitive performance with TartanVO despite being a significantly lighter model trained on only 10% of the dataset. This work establishes an end-to-end deep learning pipeline tailored for visual odometry in forested environments, leveraging forest-specific training data to optimise feature correspondence and pose estimation for improved accuracy and robustness in autonomous navigation systems.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 6","pages":"5233-5240"},"PeriodicalIF":4.6,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143848834","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":"Multi-Robot Reliable Navigation in Uncertain Topological Environments With Graph Attention Networks","authors":"Zhuoyuan Yu;Hongliang Guo;Chee-Meng Chew;Albertus Hendrawan Adiwahono;Jianle Chan;Brina Wey Tynn Shong;Wei-Yun Yau","doi":"10.1109/LRA.2025.3557751","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557751","url":null,"abstract":"This letter studies the multi-robot reliable navigation problem in uncertain topological networks, which aims at maximizing the robot team's on-time arrival probabilities in the face of road network uncertainties. The uncertainty in these networks stems from the unknown edge traversability, which is only revealed to the robot upon its arrival at the edge's starting node. Existing approaches often struggle to adapt to real-time network topology changes, making them unsuitable for varying topological environments. To address the challenge, we reformulate the problem into a Partially Observable Markov Decision Process (POMDP) framework and introduce the Dynamic Adaptive Graph Embedding method to capture the evolving nature of the navigation task. We further enhance each robot's policy learning process by integrating deep reinforcement learning with Graph Attention Networks (GATs), leveraging self-attention to focus on critical graph features. The proposed approach, namely Multi-robot Adaptive Navigation via Graph Attention-based Reinforcement learning (MANGAR) employs the generalized policy gradient algorithm to optimize the robots' real-time decision-making process iteratively. We compare the performance of MANGAR with state-of-the-art reliable navigation algorithms as well as Canadian traveller problem solutions in a range of canonical transportation networks, demonstrating improved adaptability and performance in uncertain topological networks. Additionally, real-world experiments with two robots navigating within a self-constructed indoor environment with uncertain topological structures demonstrate MANGAR's practicality.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5082-5089"},"PeriodicalIF":4.6,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856198","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}
Fan Yang;Thomas Power;Sergio Aguilera Marinovic;Soshi Iba;Rana Soltani Zarrin;Dmitry Berenson
{"title":"Multi-Finger Manipulation via Trajectory Optimization With Differentiable Rolling and Geometric Constraints","authors":"Fan Yang;Thomas Power;Sergio Aguilera Marinovic;Soshi Iba;Rana Soltani Zarrin;Dmitry Berenson","doi":"10.1109/LRA.2025.3557752","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557752","url":null,"abstract":"Parameterizing finger rolling and finger-object contacts in a differentiable manner is important for formulating dexterous manipulation as a trajectory optimization problem. In contrast to previous methods which often assume simplified geometries of the robot and object or do not explicitly model finger rolling, we propose a method to further extend the capabilities of dexterous manipulation by accounting for non-trivial geometries of both the robot and the object. By integrating the object's Signed Distance Field (SDF) with a sampling method, our method estimates contact and rolling-related variables in a differentiable manner and includes those in a trajectory optimization framework. This formulation naturally allows for the emergence of finger-rolling behaviors, enabling the robot to locally adjust the contact points. To evaluate our method, we introduce a benchmark featuring challenging multi-finger dexterous manipulation tasks, such as screwdriver turning and in-hand reorientation. Our method outperforms baselines in terms of achieving desired object configurations and avoiding dropping the object. We also successfully apply our method to a real-world screwdriver turning task and a cuboid alignment task, demonstrating its robustness to the sim2real gap.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5170-5177"},"PeriodicalIF":4.6,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143839944","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}
Kuan Yuee Tan;Neha P. Garg;Manoj Ramanathan;Wei Tech Ang
{"title":"Robust Side Following Robotic Wheelchair by Using Homotopy Class of Human Intention","authors":"Kuan Yuee Tan;Neha P. Garg;Manoj Ramanathan;Wei Tech Ang","doi":"10.1109/LRA.2025.3557304","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557304","url":null,"abstract":"A side-by-side following robot can alleviate the need for wheelchair pushing, thereby reducing the burden on caregivers and porters. Existing works on side-by-side following cannot be easily adapted to different environments as they either need prior knowledge about the human's path or the robot may take a different path than the human when moving around obstacles. In this work, we propose a side-by-side following system that ensures that the wheelchair takes the same path as the human around obstacles while avoiding them by leveraging the homotopy class of the human's intended path within a shared control framework. Our system can also be easily deployed in various real-world environments as it only needs prior knowledge of the human's final goal. Through quantitative experiments in simulation, we show that our method can perform significantly better compared to a baseline approach that does not leverage the knowledge of the human's intended path. We further validate our method through testing with human subjects using real robotic wheelchair. Our results show higher preference for our method as compared to the baseline method.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5018-5025"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143824566","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}
Sungwoon Yoon;Junyong Song;Hyun Jun Cho;Sangshin Park;Jin Tak Kim;Hyouk Ryeol Choi;Jungsan Cho
{"title":"Lower-Limb Exoskeleton Reflecting Asymmetric Movements of Femoral Condyle","authors":"Sungwoon Yoon;Junyong Song;Hyun Jun Cho;Sangshin Park;Jin Tak Kim;Hyouk Ryeol Choi;Jungsan Cho","doi":"10.1109/LRA.2025.3557310","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557310","url":null,"abstract":"The human knee has an asymmetric biomechanical structure, with the medial condyle being larger than the lateral condyle to support loads and facilitate diverse movements, thus leading to complex joint movements. These complex biomechanical properties often lead to kinematic misalignment between the knee joint and conventional wearable robots, resulting in restricted movement, unexpected forces, and frame disengagement. To address these limitations, we propose an asymmetric lower-limb exoskeleton with distinct features for the medial and lateral sides of the knee joint to suitably adapt to the knee characteristics. The exoskeleton aims to evenly distribute the load applied to the joint and naturally track its complex movements. The exoskeleton reflects the load-bearing properties of the medial side and the flexibility of the lateral side. Moreover, its frame structure supports both the medial and lateral sides to minimize the load on the knee. Fewer degrees of freedom (DOFs) are applied to the medial side to distribute the load on the joint, while additional DOFs are introduced in the lateral side for flexible movement tracking. Tendon-driven actuation assists knee motion, minimizes the joint volume and weight, and separates the joint from the actuator. Experimental results demonstrate that the proposed exoskeleton improves misalignment issues and complements the wearer's muscle strength during walking, indicating its potential for assistance and enhanced functionality.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5050-5057"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856267","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":"An Untethered Bionic Soft Robotic Vehicle With Magnetic Actuation and Control for Multi-Scenario Applications","authors":"Wenguang Yang;Zezheng Qiao;Zhizheng Gao;Haibo Yu","doi":"10.1109/LRA.2025.3557302","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557302","url":null,"abstract":"In this study, a bionic soft robotic vehicle is designed with overall dimensions of 22.7 mm in length and 22 mm in width and weighs just 1.69 g. The robotic vehicle can realize precise motion control through a rotating magnetic field generated by Helmholtz coils. In the rotating magnetic field environment, by changing the parameters of the input current waveform, the vehicle can perform a variety of modes of motion, such as forward, backward, steering, and more. Furthermore, the robotic vehicle exhibits excellent motion performance. It is capable of moving stably on a variety of surfaces, including glass, wooden board, frosted board, and sandy soil, and can even successfully climb over obstacles as high as 30°. In particular, it can climb over a 20° barrier at 11.87 mm/s when loaded with three times its own weight (5 g). At the functional level, the robotic vehicle showcases its capabilities in obstacle clearance, path planning, and detection, underscoring its potential for performing complex tasks.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 6","pages":"5289-5296"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856139","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}
Tiantian Dong;Xianlu Song;Yonghong Zhang;Xiayang Qin;Yunping Liu;Zongchun Bai
{"title":"ViT-Enabled Task-Driven Autonomous Heuristic Navigation Based on Deep Reinforcement Learning","authors":"Tiantian Dong;Xianlu Song;Yonghong Zhang;Xiayang Qin;Yunping Liu;Zongchun Bai","doi":"10.1109/LRA.2025.3557305","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557305","url":null,"abstract":"In unknown environments lacking prior maps, achieving effective visual understanding is crucial for building highly efficient task - driven autonomous navigation systems. In this paper, we propose a vision - enabled goal - oriented autonomous navigation system. This system uses a novel hybrid vision Transformer architecture as the core of its visual perception. Our approach integrates an intermediate waypoint exploration strategy, breaking down a given task into a series of consecutive subtargets. These subtargets are then fed into the scene encoder as an important part of the current physical task state, thereby achieving seamless integration of scene representation and current target information. Based on this, we utilize a deep reinforcement learning framework to develop a local navigation strategy for each subtarget. Given the challenge of addressing the sparse reward function problem, we design a novel hazardous region transfer function.In the simulation experiment stage, we validate the effectiveness of the proposed autonomous navigation system and compare it with other deep - reinforcement - learning - based navigation methods. The experimental results show that our method has significant advantages in terms of navigation success rate and efficiency. Additionally, in the Sim2Real experiments, compared with other algorithms, our method demonstrates greater robustness and mobility.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 6","pages":"5297-5304"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856272","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":"A Probabilistic Formulation of LiDAR Mapping With Neural Radiance Fields","authors":"Matthew McDermott;Jason Rife","doi":"10.1109/LRA.2025.3557301","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557301","url":null,"abstract":"In this letter we reexamine the process through which a Neural Radiance Field (NeRF) can be trained to produce novel LiDAR views of a scene. Unlike image applications where camera pixels integrate light over time, LiDAR pulses arrive at specific times. As such, multiple LiDAR returns are possible for any given detector and the classification of these returns is inherently probabilistic. Applying a traditional NeRF training routine can result in the network learning “phantom surfaces” in free space between conflicting range measurements, similar to how “floater” aberrations may be produced by an image model. We show that by formulating loss as an integral of probability (rather than as an integral of optical density) the network can learn multiple peaks for a given ray, allowing the sampling of first, <inline-formula><tex-math>$text{n}^{text{th}}$</tex-math></inline-formula>, or strongest returns from a single output channel.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 6","pages":"5409-5416"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10947591","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143860888","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Julius Rückin;David Morilla-Cabello;Cyrill Stachniss;Eduardo Montijano;Marija Popović
{"title":"Towards Map-Agnostic Policies for Adaptive Informative Path Planning","authors":"Julius Rückin;David Morilla-Cabello;Cyrill Stachniss;Eduardo Montijano;Marija Popović","doi":"10.1109/LRA.2025.3557233","DOIUrl":"https://doi.org/10.1109/LRA.2025.3557233","url":null,"abstract":"Robots are frequently tasked to gather relevant sensor data in unknown terrains. A key challenge for classical path planning algorithms used for autonomous information gathering is adaptively replanning paths online as the terrain is explored given limited onboard compute resources. Recently, learning-based approaches emerged that train planning policies offline and enable computationally efficient online replanning performing policy inference. These approaches are designed and trained for terrain monitoring missions assuming a single specific map representation, which limits their applicability to different terrains. To address this limitation, we propose a novel formulation of the adaptive informative path planning problem unified across different map representations, enabling training and deploying planning policies in a larger variety of monitoring missions. Experimental results validate that our novel formulation easily integrates with classical non-learning-based planning approaches while maintaining their performance. Our trained planning policy performs similarly to state-of-the-art map-specifically trained policies. We validate our learned policy on unseen real-world terrain datasets.","PeriodicalId":13241,"journal":{"name":"IEEE Robotics and Automation Letters","volume":"10 5","pages":"5114-5121"},"PeriodicalIF":4.6,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143856279","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}