arXivSub Start free trial

IROS 2025 Papers — Page 19

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1984 papers

ThermalLoc: A Vision Transformer-Based Approach for Robust Thermal Camera Relocalization in Large-Scale Environments

Yu Liu, Changhao Chen

Pose EstimationConvolutional Neural NetworkTransformerSimultaneous Localization and MappingImage

🎯 What it does: Designed an end-to-end deep learning method called ThermalLoc for thermal imaging camera relocalization.

Three-Dimensional Anatomical Data Generation Based on Artificial Neural Networks*

Ann-Sophia Müller, T. Qiu

SegmentationGenerationData SynthesisGenerative Adversarial NetworkMeshBiomedical DataUltrasound

🎯 What it does: Developed a physics-based automated workflow for generating 3D anatomical data, utilizing artificial prostate models and ultrasound scans to train neural networks for image segmentation and 3D mesh reconstruction, providing surgical simulation and performance feedback.

Three-DOF controlled flight in palm-scale micro robotic blimp driven by flapping wings

Jie Chen, Xuezhong Wu

Robotic Intelligence

🎯 What it does: Designed and implemented a piezoelectric-driven micro balloon robot using two flapping-wing propulsion units to achieve three-degree-of-freedom (3-DOF) control, along with wireless remote control, attitude sensing, and closed-loop yaw control.

Throwing Planning Diffusion: A Solution to Learning and Planning of Robotic Throwing

Ziqi Xu, Xuechao Duan

OptimizationRobotic IntelligenceDiffusion model

🎯 What it does: Combines diffusion models with model-free throwing methods, utilizing backward reachable tubes to search for throwing configurations, sampling from posterior trajectory distributions conditioned on configurations, and employing multiple trajectory optimization methods to generate feasible, smooth, and collision-free throwing trajectories.

Thruster-Enhanced Locomotion: A Decoupled Model Predictive Control with Learned Contact Residuals

Chenghao Wang, Alireza Ramezani

Robotic Intelligence

🎯 What it does: Proposes separating leg Raibert-type control from model predictive control (MPC) thrust control, and enhancing the stability of thrust-assisted narrow-path walking by compensating for leg-ground impact through learned contact residual dynamics (CRD).

TIETracker: A CLIP-based RGB-T Tracking via Feature Interaction and Semantic Enhancement

Weidai Xia, Fangfang Li

Object TrackingRepresentation LearningTransformerVision Language ModelContrastive LearningMultimodality

🎯 What it does: Proposed a CLIP-based RGB-T tracking algorithm called TIETracker, which guides the network to learn multi-modal target representations and promotes multi-modal feature interaction through textual information.

Time-Optimal Path Parameterization with Viscous Friction and Jerk Constraints based on Reachability Analysis

Maximilian Dio, Andreas Völz

OptimizationRobotic Intelligence

🎯 What it does: Proposed a time-optimal path parameterization method based on reachability analysis for robotic systems with viscous friction and impact constraints.

Time-Optimal Trajectory Generation with Multi-level Continuous Kinodynamics Constraints

Ruixuan Liu, J. Leu

OptimizationRobotic Intelligence

🎯 What it does: Proposes a time-optimal trajectory generation method called TOTG-C, considering instantaneous kinematic (IKD) and multi-level continuous kinematic (CKD) constraints.

Tiny LiDARs for Manipulator Self-Awareness: Sensor Characterization and Initial Localization Experiments

Giammarco Caroleo, P. Maiolino

Robotic IntelligencePoint Cloud

🎯 What it does: Using a miniature VL53L5CX ToF sensor to acquire a rough point cloud, experimental calibration is conducted to address the sensor's dependency on readings, relative distance, and direction. A probabilistic sensor model is proposed and validated for target object pose estimation within a particle filter.

TK-Planes: Tiered K-Planes with High Dimensional Feature Vectors for Dynamic UAV-based Scenes

Christopher Maxey, Heesung Kwon

GenerationNeural Radiance FieldVideo

🎯 What it does: Propose an extended K-Planes Neural Radiance Field (NeRF) method that enhances the neural rendering quality of dynamic scenes captured by unmanned aerial vehicles (UAVs) by storing hierarchical high-dimensional feature vectors.

TopoLiDM: Topology-Aware LiDAR Diffusion Models for Interpretable and Realistic LiDAR Point Cloud Generation

Jiuming Liu, Hesheng Wang

GenerationData SynthesisAutonomous DrivingExplainability and InterpretabilityGraph Neural NetworkDiffusion modelAuto EncoderPoint Cloud

🎯 What it does: Designed and implemented the TopoLiDM framework for generating high-fidelity LiDAR point clouds, combining graph neural networks with diffusion models and incorporating topological regularization.

Topology-Driven Trajectory Optimization for Modelling Controllable Interactions Within Multi-Vehicle Scenario

Chang Ma, Wenchao Ding

Autonomous DrivingOptimization

🎯 What it does: Propose a differentiable local homotopy-invariant metric for multi-vehicle trajectory optimization, generating controllable interactive trajectories.

TOPP-DWR: Time-Optimal Path Parameterization of Differential-Driven Wheeled Robots Considering Piecewise-Constant Angular Velocity Constraints

Yong Li, Hui Cheng

OptimizationRobotic Intelligence

🎯 What it does: Proposed a time-optimal path parameterization algorithm named TOPP-DWR, applicable to differential drive wheeled robots, considering piecewise constant angular velocity constraints as well as joint velocity, linear velocity, and linear acceleration constraints.

ToSA: Token Merging with Spatial Awareness

Hsiang-Wei Huang, Jenq-Neng Hwang

Computational EfficiencyRepresentation LearningTransformerImage

🎯 What it does: Propose a Token merging method (ToSA) that combines semantic and spatial awareness to accelerate Vision Transformers.

Touch-Linked Sleeve: A Haptic Interface for Augmented Tactile Perception in Robotic Teleoperation

Yatao Leng, Chenxi Xiao

Robotic Intelligence

🎯 What it does: Proposed and implemented Touch-Linked Sleeve (TLS), mapping tactile information from the robot arm to the operator's skin to enhance tactile perception during teleoperation

Touch-Sensitive Hand Interactions for Social Robots Using Fiber Bragg Grating Sensors

María Gaitán-Padilla, Camilo A. R. Diaz

ClassificationRobotic IntelligenceTabular

🎯 What it does: Developed and validated a fiber Bragg grating (FBG) sensor network integrated into the hand of the CASTOR robot for classifying complex physical human-robot interactions.

Toward Dynamic Control of Tendon-driven Continuum Robots using Clarke Transform

Christian Muhmann, J. Burgner-Kahrs

Robotic Intelligence

🎯 What it does: Proposes a dynamic model and control framework for multi-segment, multi-tendon-driven continuum robots, and implements linear and constraint-based controllers on a 2D manifold.

Towards Accurate Brain Electrode Implantation via Cross-modality Fusion of White-light and Photoacoustic Microscopy

Yuxuan Liu, Guang-Zhong Yang

Robotic IntelligenceConvolutional Neural NetworkMultimodalityBiomedical Data

🎯 What it does: Fusion of white light microscopy with preoperative photoacoustic microscopy to enhance visibility of cerebral surface microvasculature, and application of the fusion results to robotic system planning and operation for minimally invasive brain electrode implantation; meanwhile, a multi-modal data preprocessing workflow and a 2.5D fusion network with depth encoding were proposed;

Towards an Extremely Robust Baby Robot With Rich Interaction Ability for Advanced Machine Learning Algorithms

Mohannad Alhakami, Jürgen Schmidhuber

Robotic IntelligenceReinforcement Learning

🎯 What it does: Designed and verified a semi-soft structured, full-camera non-contact sensor robot limb with easily replaceable failure points to support long-term learning tasks of reinforcement learning algorithms on physical prototypes.

Towards Autonomous Indoor Parking: A Globally Consistent Semantic SLAM System and A Semantic Localization Subsystem

Yichen Sha, Hesheng Wang

Autonomous DrivingSimultaneous Localization and MappingMultimodality

🎯 What it does: Proposed a globally consistent semantic SLAM system GCSLAM and a semantic fusion localization subsystem SF-Loc, achieving precise semantic mapping and robust localization within parking lots.

Towards Autonomous Robotic Electrosurgery via Thermal Imaging

Naveed D. Riaziat, Jeremy D. Brown

OptimizationRobotic IntelligenceImage

🎯 What it does: Developed the ThERMO system, which uses thermal imaging feedback to adaptively control the speed of the electrosurgical device, in order to reduce thermal injury and balance cutting force

Towards Data-Driven Adaptive Exoskeleton Assistance for Post-stroke Gait

F. Weigend, C. J. Walsh

Robotic IntelligenceConvolutional Neural NetworkTime SeriesBiomedical Data

🎯 What it does: A data-driven adaptive torque estimation model for ankle flexion/extension assistance in post-stroke gait was developed, and a wearable prototype was implemented.

Towards Deformation Modeling and Simulation of a Soft and Inflatable Endoscopic Vision-Based Tactile Sensing Balloon for Cancer Diagnosis

Ozdemir Can Kara, F. Alambeigi

OptimizationRobotic IntelligenceBiomedical Data

🎯 What it does: Proposed and validated a simulation model based on SOFA software to analyze the deformation behavior of E-VTSB under different pressures, with experimental validation of its accuracy.

Towards Design and Development of a Concentric Tube Steerable Drilling Robot for Creating S-shape Tunnels for Pelvic Fixation Procedures

Y. Kulkarni, F. Alambeigi

Robotic IntelligenceBiomedical Data

🎯 What it does: Designed and developed a 4-degree-of-freedom pelvic coaxial tube controllable drill system (pelvic CT-SDR), capable of creating S-shaped drilling paths that conform to natural anatomical curvature in the pelvis, with its performance evaluated through multiple S-shaped drilling experiments on simulated bone models.

Towards Efficient Image-goal Navigation: A Self-supervised Transformer-based Reinforcement Learning Approach

Qizhen Weng, Xiangwei Zhu

TransformerReinforcement LearningImage

🎯 What it does: Proposes a self-supervised Transformer-based reinforcement learning method for image goal navigation, which utilizes a dual attention shared Transformer to predict masked visual-action embeddings and generate policies, thereby fully leveraging spatiotemporal relationships in visual-action history.

Towards Emotion Co-regulation with LLM-powered Socially Assistive Robots: Integrating LLM Prompts and Robotic Behaviors to Support Parent-Neurodivergent Child Dyads

Jing Li, Emilia I. Barakova

Robotic IntelligenceTransformerLarge Language ModelPrompt EngineeringTextAudio

🎯 What it does: Developed a social assistive robot (MiRo-E) that integrates LLM prompts with robotic behavior, and conducted preliminary testing and qualitative analysis with two parent-neurodiverse child pairs, aiming to support emotion co-regulation.

Towards Extrinsic Dexterity Grasping in Unrestricted Environments

Chengzhong Ma, Nanning Zheng

Robotic IntelligenceVision Language ModelVision-Language-Action ModelDiffusion model

🎯 What it does: Proposed the ExDiff method, which utilizes Vision-Language Models (VLM) to perceive the environment and generate instructions, and then employs the Goal-Conditioned Action Diffusion (GCAD) model to predict low-level action sequences, achieving dexterous grasping of non-graspable objects.

Towards Fully Autonomous Robotic Ultrasound-guided Biopsy for Superficial Organs

Chenwei Wang, Zhenglong Sun

Robotic IntelligenceImageBiomedical DataUltrasound

🎯 What it does: Proposed a fully autonomous robotic operation framework for ultrasound-guided biopsy of superficial organs, integrating real-time slice-to-volume registration, navigation, and needle insertion mechanisms to complete the entire biopsy process according to the operational protocol.

Towards Open-World Human Action Segmentation Using Graph Convolutional Networks

Hao Xing, Gordon Cheng

SegmentationData SynthesisGraph Neural NetworkVideo

🎯 What it does: Proposed a framework for open-world human action segmentation, evaluated on two datasets.

Towards Physically Realizable Adversarial Attacks in Embodied Vision Navigation

Meng Chen, Jianqin Yin

Adversarial Attack

🎯 What it does: A physically realizable adversarial attack method for embodied visual navigation was studied and implemented, proposing attack patches with learnable opacity and texture, and employing multi-view optimization and two-stage opacity optimization.

Towards Robust Sensor-Fusion Ground SLAM: A Comprehensive Benchmark and A Resilient Framework

Deteng Zhang, Jie Yin

Autonomous DrivingSimultaneous Localization and MappingImageMultimodalityPoint CloudBenchmark

🎯 What it does: This paper introduces the M3DGR dataset, evaluates forty SLAM systems, and develops the Ground-Fusion++ fusion framework.

Towards Safe Imitation Learning via Potential Field-Guided Flow Matching

Haoran Ding, Yoshihiko Nakamura

Safty and PrivacyFlow-based Model

🎯 What it does: Proposes a potential field guided flow matching strategy called PF2MP, which can achieve safe imitation learning in both simulation and real environments.

Towards Safe Reinforcement Learning with Reduced Conservativeness: A Case Study on Drone Flight Control

Loizos Hadjiloizou, Danica Kragic

Safty and PrivacyRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a framework that refines the disturbance model through online data collection and uses zonotopic reachability analysis to evaluate the safety of learning-based controllers, reducing the conservatism of formal methods during real UAV canyon flights.

Towards Surgical Task Automation: Actor-Critic Models Meet Self-Supervised Imitation Learning*

Jingshuai Liu, S. Tsaftaris

Robotic IntelligenceReinforcement LearningBiomedical Data

🎯 What it does: Propose an RL framework AC-SSIL based on Actor-Critic, which integrates self-supervised imitation learning (SSIL) to introduce expert demonstrations containing only states into RL.

Towards the Benchmarking of Embodied Sensors for Pose Tracking in Octopus-inspired Robotic Arms

Michele Martini, Barbara Mazzolai

Pose EstimationRobotic IntelligenceTime SeriesBenchmark

🎯 What it does: Benchmarking two types of proprioceptive sensors embedded in a flexible soft robotic arm (FBG optical sensors and IMU systems) to evaluate their applicability in air and underwater environments.

Towards Zero-Knowledge Task Planning via a Language-based Approach

Liam Merz Hoffmeister, Daniel Rakita

Robotic IntelligenceTransformerLarge Language ModelText

🎯 What it does: Proposed and formalized the Zero-Knowledge Task Planning (ZKTP) problem, leveraging large language models (LLMs) to decompose natural language instructions into subtasks and generate behavior trees (BTs) for execution; if errors occur during execution, the LLM can dynamically adjust the BT to achieve iterative optimization.

TR-LLM: Integrating Trajectory Data for Scene-Aware LLM-Based Human Action Prediction

Kojiro Takeyama, Misha Sra

Large Language ModelMultimodalityTime Series

🎯 What it does: Propose a multimodal human action prediction framework that integrates trajectory data with large language models (LLMs).

TRACE: A Self-Improving Framework for Robot Behavior Forecasting with Vision-Language Models

Gokul Puthumanaillam, M. Ornik

Robotic IntelligenceChain-of-Thought

🎯 What it does: Proposes the TRACE framework, which leverages visual language models (VLM) and tree-of-thoughts generation combined with metric counterfactual exploration to iteratively improve robot behavior trajectory prediction.

TRACER: Thrust Auto-calibration and Ground Effect Estimation Using Onboard Force Sensitive Resistor Array for Multirotors

Baichuan Lou, Chen Lv

Robotic Intelligence

🎯 What it does: Propose a low-cost onboard force-sensitive resistor array and a unified algorithm to achieve automatic calibration of the thrust coefficients and estimation of the ground effect for multirotor drones.

Tracking Any Point with Frame-Event Fusion Network at High Frame Rate

Jiaxiong Liu, Dewen Hu

Object TrackingTransformerMultimodality

🎯 What it does: Proposed a point tracker FE-TAP that combines image frames with events, achieving high frame rate and robust point tracking.

Tracking Control of 7-DOF Redundant Manipulators with Enhanced Null Space Compliance

Xinyang Tian

Robotic Intelligence

🎯 What it does: A tracking control method for a 7-degree-of-freedom redundant manipulator is proposed, capable of precisely executing time-varying pose trajectories and achieving limited compliant behavior within the null space.

Tracking Highly Dynamic Humanoid Motion with Dynamic IMU Measurement Fusion

Jeronimo Cox, Tomonari Furukawa

Pose EstimationTime Series

🎯 What it does: Propose a method utilizing IMU measurements to achieve high dynamic human motion tracking, using accelerometers to correct rotational rates;

Tracking-Aware Deformation Field Estimation for Non-rigid 3D Reconstruction in Robotic Surgeries

Zeqing Wang, Yutong Ban

Robotic IntelligenceNeural Radiance FieldOptical FlowVideoMeshBiomedical Data

🎯 What it does: Proposed and implemented a Tracking-Aware Deformation Field (TADF) framework for non-rigid 3D reconstruction and deformation estimation of soft tissues in robotic surgery.

Training People to Reward Robots

Endong Sun, Matthew Howard

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Utilize machine teaching to guide beginners in improving teaching skills within an RLfD environment, and experimentally verify its enhancement of robot learning performance

Trajectory Generation for Humanoid Backflips and Jumps Based on Whole-Body Dynamics Optimization with Consideration of KKT Residual Convergence

Masanori Konishi, Kei Okada

Optimization

🎯 What it does: Propose a two-phase optimization strategy based on whole-body dynamics optimization to generate trajectories for whole-body jumping motions (including backflips, front flips, yaw twists, etc.).

Trajectory Optimization for In-Hand Manipulation with Tactile Force Control

Haegu Lee, Christoffer Sloth

OptimizationRobotic Intelligence

🎯 What it does: A trajectory optimization framework based on nonlinear programming is proposed for a robotic hand equipped with magnetic tactile sensors, enabling the fingertips to change contact points on the finger geometry to achieve rolling manipulation of objects.

TrajFlow: Multi-modal Motion Prediction via Flow Matching

Qi Yan, Renjie Liao

Autonomous DrivingFlow-based ModelTime SeriesSequential

🎯 What it does: Developed the TrajFlow framework, which employs flow matching technology to generate multiple feasible future trajectories in a single inference process, significantly reducing computational overhead.

Transfer Learning for Walking Speed Estimation Across Novel Prosthetic Devices and Populations

Jairo Y. Maldonado-Contreras, Aaron J. Young

Domain AdaptationRepresentation LearningConvolutional Neural NetworkSupervised Fine-TuningBiomedical Data

🎯 What it does: Proposed a transfer learning framework for gait speed estimation across new prosthetic devices and populations, pre-training a CNN on publicly available AB and OSL prosthetic data, then fine-tuning with limited PK prosthetic data.

Transferable Latent-To-Latent Locomotion Policy for Efficient and Versatile Motion Control of Diverse Legged Robots

Ziang Zheng, S. Li

Robotic IntelligenceSupervised Fine-TuningReinforcement LearningDiffusion modelAuto Encoder

🎯 What it does: Propose a transferable latent space motion policy framework, pre-training latent-to-latent motion policies combined with multi-task specific observation encoders and action decoders, utilizing a diffusion recovery module to retain key information; during the finetuning phase, fix the motion policy and only optimize the lightweight encoder and decoder to achieve rapid adaptation.

Transferring Kinesthetic Demonstrations across Diverse Objects for Manipulation Planning

Dibyendu Das, I. V. Ramakrishnan

Robotic Intelligence

🎯 What it does: Propose a method that identifies key positions and associates motion transfer frameworks with manipulated objects and targets, tracks and transfers the paths of these frameworks to achieve motion planning transfer for objects with different geometries.

Transformable Modular Robots: A CPG-Based Approach to Independent and Collective Locomotion

Jiayu Ding, Zhenyu Gan

Robotic Intelligence

🎯 What it does: Proposed a modular robot system where each module has independent drive, battery, and control, utilizing a hierarchical central pattern generator (CPG) to achieve autonomous movement of individual modules and coordinated collaboration, supporting seamless transitions from single-module to multi-module configurations;

Transformer-based Motion Model for Robust Target Tracking under Intermittent and Noisy Measurements

Andres Pulido, Jane Shin

Object TrackingTransformer

🎯 What it does: Propose multiple Transformer-based motion models for learning target dynamics under noisy and intermittent measurements, integrated with particle filters and information-driven planners to achieve target tracking

Transformer-Based Multi-Agent Reinforcement Learning Method With Credit-Oriented Strategy Differentiation

Kaixuan Huang, Ziqi Wei

TransformerReinforcement LearningBenchmark

🎯 What it does: Propose a Transformer-based multi-agent reinforcement learning method called TMRC

Transformer-Based Spatio-Temporal Association of Apple Fruitlets

Harry Freeman, George Kantor

Object TrackingTransformerImageAgriculture Related

🎯 What it does: Proposed a Transformer-based method for achieving spatiotemporal association of apple grains in stereo images under different dates and camera poses.

Transporting Heavy Payloads with a Humanoid riding a Hoverboard

Simon Armleder, Gordon Cheng

Robotic Intelligence

🎯 What it does: Developed a control system that enables a humanoid robot to ride a two-wheeled hoverboard for rapid transportation, and stabilize large cargo through full-body grasping (chest and arms).

TransSoft: The Low-Cost, Adaptable, and Radial Reconfigurable Soft Hand for Diverse Object Grasping

Yongchong Gu, Yanwei Fu

Robotic Intelligence

🎯 What it does: Developed a low-cost, adaptable, radially reconfigurable soft robotic hand called TransSoft, which can dynamically adjust its grasping strategy based on the object's size, shape, and material.

Triple-S: A Collaborative Multi-LLM Framework for Solving Long-Horizon Implicative Tasks in Robotics

Zixi Jia, Qinghua Liu

Robotic IntelligenceLarge Language Model

🎯 What it does: Proposes a multi-LLM collaborative Triple-S framework to address robot long-horizon implication tasks, improving success rates through a closed-loop Simplification-Solution-Summary process.

TTTFusion: A Test-Time Training-Based Strategy for Multimodal Medical Image Fusion in Surgical Robots

Qinhua Xie, Hao Tang

Robotic IntelligenceMultimodalityBiomedical Data

🎯 What it does: Proposed a multimodal medical image fusion strategy called TTTFusion based on test-time training (TTT) for real-time fusion of medical images in surgical robots

TVFET-VD:Time-Varying Formation Encircling and Tracking Control Based on Visual Detection

Guang Yang, Yuan Ping

Object DetectionObject TrackingConvolutional Neural NetworkImage

🎯 What it does: Proposed a complete process for multi-rotor drones from target detection, localization to surrounding tracking;

TWC-SLAM: Multi-Agent Cooperative SLAM with Text Semantics and WiFi Features Integration for Similar Indoor Environments

Chunyu Li, Qingquan Li

OptimizationRobotic IntelligenceSimultaneous Localization and MappingTextMultimodalityPoint Cloud

🎯 What it does: This paper proposes the TWC-SLAM framework, which integrates text semantics and WiFi signal features for multi-robot collaborative SLAM in similar indoor environments, achieving localization identification, loop closure detection, cross-robot point cloud alignment, and global optimization.

TwinTac: A Wide-Range, Highly Sensitive Tactile Sensor with Real-To-Sim Digital Twin Sensor Model

Xiyan Huang, Chenxi Xiao

ClassificationData SynthesisReinforcement LearningPhysics Related

🎯 What it does: Proposed the TwinTac system, which combines high-sensitivity, wide-measurement-range physical tactile sensors with their digital twin models for data generation in cross-domain learning and reinforcement learning scenarios;

Two-dimensional Trajectory Tracking of a Magnetic Continuum Robot by Optimal Magnet Manipulation

Lijun Hao, Zhe Zheng

OptimizationRobotic Intelligence

🎯 What it does: Propose a 2D trajectory tracking scheme for magnetic continuous robots based on magnetic field manipulation, and fabricate a prototype to verify its feasibility.

Two-stage Learning Framework Combining Joint-level Reinforcement Learning and Muscle-level Adaptation for Musculoskeletal Locomotion

Laurie Azoulay, M. Hayashibe

Robotic IntelligenceRecurrent Neural NetworkReinforcement Learning

🎯 What it does: Propose a two-stage learning framework that combines a reinforcement learning (RL) joint controller with an LSTM muscle controller to plan body movements and compute muscle forces, simplifying the learning of joint and muscle redundancies while enhancing interpretability.

U-Snake: A Small-Sized Smart Underwater Snake Robot

Bowen Wang, Changhong Fu

Robotic Intelligence

🎯 What it does: Designed and implemented a small, lightweight intelligent underwater snake-like robot named U-Snake, and developed a curvature-based path following method and an integrated controller for it.

UAV Chain Network Creation in Cluttered Environments with Flocking Rules and Routing Data

Théotime Balaguer, Isabelle Fantoni

Robotic IntelligencePhysics Related

🎯 What it does: Proposed a distributed UAV link network construction method based on an adaptive swarm flying model.

UAV See, UGV Do: Aerial Imagery and Virtual Teach Enabling Zero-Shot Ground Vehicle Repeat

Desiree Fisker, Tim D. Barfoot

Autonomous DrivingNeural Radiance FieldImagePoint CloudMesh

🎯 What it does: Proposed the VirT&R framework, which uses drone aerial images to train NeRF for generating high-precision point clouds and textured meshes, plans paths in simulated environments, and employs NeRF submaps with LiDAR Teach & Repeat (LT&R) for zero-shot, GPS-deprived navigation of unmanned ground vehicles in real environments.

UAV Video Deblurring via Motion-Aware Diffusion: A Path to Robust Target Detection

Zhiqiang Hu, Masatoshi Ishikawa

RestorationObject DetectionDiffusion modelVideo

🎯 What it does: Propose an efficient defocusing method applicable to UAV videos.

UAV-DETR: Efficient End-to-End Object Detection for Unmanned Aerial Vehicle Imagery

Huaxiang Zhang, Guo-Niu Zhu

Object DetectionTransformerImage

🎯 What it does: Proposed an efficient end-to-end detection transformer framework for UAV imagery, UAV-DETR, which includes multi-scale feature fusion and frequency enhancement modules, frequency-focused downsampling modules, and semantic alignment and calibration modules.

UAV-MaLO: Mamba-Augmented YOLO Hybrid Architecture for UAV Micro-Object Detection in Autonomous Robotics

Lennox Wei, Shaohui Liu

Object DetectionRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Proposes UAV-MaLO, a micro-target detection framework that integrates state-space modeling principles into YOLO.

UDSH: An Unsupervised Deep Image Stitching and De-Occlusion Method for Heavy Occlusion Scene

Kaixin Chen, Mengyin Fu

Image HarmonizationRestorationImage

🎯 What it does: Proposes an unsupervised depth image stitching and occlusion removal method for heavy occlusion scenarios.

UF-RNN: Real-Time Adaptive Motion Generation Using Uncertainty-Driven Foresight Prediction

Hyogo Hiruma, Tetsuya Ogata

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Proposed an Uncertainty-driven Foresight Recurrent Neural Network (UF-RNN), which adjusts hidden states by internally simulating multiple future trajectories and minimizing prediction variance, enabling adaptive exploration of actions in high-uncertainty environments, and evaluates the door-opening task in simulation and real robots.

ULRVT II: A Novel Upper Limb Rehabilitation Robot with Joint Synergy Control and Evaluation for Virtual Training*

Lei Yang, Yili Fu

Robotic Intelligence

🎯 What it does: Designed and tested an upper limb rehabilitation robot for virtual training (ULRVT II), and built a rehabilitation platform and assessment system.

Ultra-Wideband assisted Visual-Inertial Localization Correction System with Position-Unknown UWB Anchors

Yu Xing, Xiaoxue Feng

OptimizationSimultaneous Localization and Mapping

🎯 What it does: Proposes a correction system that utilizes UWB-assisted visual inertial odometry (VIO), and designs a lightweight initialization and joint estimation method to correct VIO drift under unknown UWB anchor positions.

UltraDP: Generalizable Carotid Ultrasound Scanning with Force-Aware Diffusion Policy

Ruoqu Chen, Xiang Li

Safty and PrivacyRobotic IntelligenceDiffusion modelMultimodalityUltrasound

🎯 What it does: Proposed the UltraDP method based on diffusion strategy, which can achieve autonomous carotid ultrasound scanning under multi-modal inputs, and position the artery at the image center through a dedicated guidance module.

UltraTac: Integrated Ultrasound-Augmented Visuotactile Sensor for Enhanced Robotic Perception

Junhao Gong, Wenbo Ding

Robotic IntelligenceMultimodalityUltrasound

🎯 What it does: Developed an UltraTac sensor that integrates ultrasound and visual-tactile imaging, achieving the unification of two perceptual modes into a single structure through a coaxial optoacoustic architecture;

Uncertain-aware Informative Task Planning and Assignment for Multiple-UUVs Cooperative Underwater Exploration

Chengfeng Jia, Rong Su

OptimizationPhysics Related

🎯 What it does: Proposes an uncertainty-aware multi-UUV collaborative underwater exploration framework that integrates task planning, task allocation, and prior belief updates, achieving dynamic interest region selection and allocation.

Uncertainty-Aware Knowledge Distillation for Compact and Efficient 6DoF Pose Estimation

Nassim Ali Ousalah, D. Aouada

Pose EstimationKnowledge DistillationImage

🎯 What it does: Proposes an end-to-end knowledge distillation framework based on uncertainty for lightweight keypoint-based 6DoF pose estimation.

Uncertainty-aware Motion Planning based on Stochastic Forward/Inverse Kinematics Models for Tensegrity Manipulators

Yuhei Yoshimitsu, Shuhei Ikemoto

Robotic Intelligence

🎯 What it does: Propose a motion planning method based on stochastic neural networks learning forward and inverse kinematics, utilizing network uncertainty to balance shape and stiffness

Uncertainty-Aware Multi-Robot Flocking via Learned State Estimation and Control Barrier Functions

Mattia Catellani, Lorenzo Sabattini

Robotic Intelligence

🎯 What it does: Propose a decentralized multi-robot cluster control method that utilizes state and uncertainty estimates of undetected robots to achieve collaboration without information exchange.

Uncertainty-aware Planning with Inaccurate Models for Robotized Liquid Handling

Marco Faroni, Paolo Rocco

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose an uncertainty-aware Monte Carlo Tree Search (MCTS) algorithm to address the problem of inaccurate models in robotic liquid handling.

Uncertainty-Aware Shared Control for Vision-Based Micromanipulation

Huanyu Tian, Christos Bergeles

OptimizationRobotic Intelligence

🎯 What it does: Proposed an uncertainty-aware shared control and calibration method for micro-operations, utilizing digital microscopes and multi-joint robotic arms to integrate real-time human intervention with visual-motor strategies;

Understanding and Imitating Human-Robot Motion with Restricted Visual Fields

Maulik Bhatt, Negar Mehr

Robotic IntelligenceReinforcement Learning from Human Feedback

🎯 What it does: Studied the agent's perception capability under limited field of view and range, decoupling perception models from motion strategies, and leveraging human perception modeling to better predict behavior; validated human navigation strategies within limited observation spaces through user experiments, enabled robots to learn this strategy for real-time navigation with minimal collisions, and successfully demonstrated it on physical hardware vehicles.

Understanding the Impact of Modeling Abstractions on Motion Planning for Deformable Linear Objects

Jimmy Envall, Stelian Coros

Robotic Intelligence

🎯 What it does: Studied the impact of model abstraction levels on motion planning for flexible linear objects, and explored schemes to enhance robustness and performance by removing compressive resistance and penalizing compressed states in planning objectives.

Underwater Exosuit Actuator Design for Unrestricted Bidirectional Hip Assistance During Flutter Kicking

Xiangyang Wang, Xinyu Wu

Robotic Intelligence

🎯 What it does: Designed an exoskeleton actuator that provides bidirectional hip assistance for divers performing undulating kicks underwater.

Underwater Remote Intervention Based on Satellite Communication

Xuejiao Yang, LinghanMeng

Robotic Intelligence

🎯 What it does: Proposed transferring the remote control environment from a maritime support vessel to land via satellite communication links, designed a cross-domain underwater intervention layered control architecture and shared control strategy, and verified its effectiveness in field experiments.

Underwater target 6D State Estimation via UUV Attitude Enhance Observability

Fen Liu, Rong Su

Pose Estimation

🎯 What it does: A framework for 6D relative state estimation of non-cooperative targets using a single unmanned underwater vehicle (UUV) with continuous noise distance measurements from two single-beam sonars is proposed, along with the design of observability-enhanced attitude control and Lyapunov-based trajectory tracking control;

Underwater Target Tracking with Unknown Maneuver by Remotely Operated Vehicles: A Digital Twin-Driven Strategy

Tianyi Zhang, Xinping Guan

Object TrackingReinforcement Learning

🎯 What it does: Developed a digital twin (DT)-based tracking strategy for unknown maneuvering targets, utilizing a remotely operated vehicle (ROV) to achieve target tracking; constructed a DT framework based on state prediction, and estimated the unknown state transition matrix of the target using neural networks; designed a reinforcement learning (RL) tracking controller in the DT model based on the predicted target state, applying the optimal tracking strategy to the real ROV; reduced the matching error between virtual and real ROVs by using RL optimization algorithms to leverage data interaction between the DT model and the ROV; decreased communication energy consumption and reduced dependence on the target maneuver model by periodically feeding real ROV data back to the DT model.

Uni-Zipper: A Multi-modal Perception Framework of Deformable Objects with Unpaired Data

Q. Xie, Bin He

Robotic IntelligenceTransformerVision Language ModelContrastive LearningMultimodality

🎯 What it does: Proposed a scalable multi-modal fusion framework called Uni-Zipper for enabling multi-modal perception in robots when handling deformable objects.

Unidirectional Point-Voxel Fusion for Enhanced 3D Single Object Tracking

Yuyu Jiang, Ying Yao

Object TrackingTransformerPoint Cloud

🎯 What it does: Proposes UTracker, which builds a bridge between point and voxel tracking features through unidirectional point-voxel fusion, achieving complementary enhancement via Template Enhanced Unidirectional Attention (TEUA), Historical Template Fusion (HTF), and Point-Guided Adaptive Feature Transformer (PGAFT).

Unified Locomotion Transformer with Simultaneous Sim-to-Real Transfer for Quadrupeds

Dikai Liu, Simon See

Knowledge DistillationRobotic IntelligenceTransformerReinforcement Learning

🎯 What it does: Propose a unified quadrupedal robot motion transformer (ULT) that integrates knowledge transfer and policy optimization into a single network, enabling zero-shot deployment;

UniLegs: Universal Multi-Legged Robot Control through Morphology-Agnostic Policy Distillation

Weijie Xi, Guyue Zhou

Knowledge DistillationRobotic IntelligenceTransformer

🎯 What it does: Proposes a two-stage policy distillation method based on a teacher-student framework, distilling specialized control policies for various multi-legged robot morphologies into a single Transformer student policy, thereby achieving general control for multi-legged robot morphologies.

UniTac-NV: A Unified Tactile Representation For Non-Vision-Based Tactile Sensors *

Jian Hou, Adam J. Spiers

Domain AdaptationRepresentation LearningRobotic IntelligenceAuto EncoderTime Series

🎯 What it does: Proposed an encoder-decoder architecture to unify data from non-visual tactile sensors and enable cross-sensor data transfer.

Unsupervised Anomaly Detection Improves Imitation Learning for Autonomous Racing

Yuang Geng, Ivan Ruchkin

Anomaly DetectionAutonomous DrivingConvolutional Neural NetworkAuto EncoderVideo

🎯 What it does: Propose an unsupervised anomaly detection method to automatically filter abnormal images in driving datasets, thereby improving imitation learning performance.

Unsupervised Liver Deformation Correction Network Using Optimal Transport for Image-Guided Liver Surgery

Mingyang Liu, Zhe Min

OptimizationBiomedical Data

🎯 What it does: Proposed an unsupervised liver deformation correction method called LCNet for image-guided liver surgery;

Unveiling the Potential of Segment Anything Model 2 for RGB-Thermal Semantic Segmentation with Language Guidance

Jiayi Zhao, Kailun Yang

SegmentationTransformerVision Language ModelMultimodalityBenchmark

🎯 What it does: Proposes SHIFNet, a RGB-thermal semantic segmentation framework based on Segment Anything Model 2

User Experience Estimation in Human-Robot Interaction via Multi-Instance Learning of Multimodal Social Signals

Ryo Miyoshi, Jun Baba

Robotic IntelligenceTransformerMultimodality

🎯 What it does: Propose a user experience estimation method based on multimodal social signals, utilizing a multi-instance learning framework to capture short-term and long-term interaction patterns;

Using High-Level Patterns to Estimate How Humans Predict a Robot will Behave

Sagar Parekh, Dylan P. Losey

Autonomous DrivingRobotic Intelligence

🎯 What it does: Developed a second-order mind method enabling robots to estimate how humans predict their behavior, using a discrete latent space to embed human-robot trajectories and generate high-level behavior predictions.

Using Upper Limb Carrying Exoskeleton with Dual-Model Torque Control Strategy to Reduce Load Impact

Daming Liu, Yanhe Zhu

Robotic IntelligenceRecurrent Neural NetworkTransformer

🎯 What it does: Proposed a dual-model multimodal fusion control strategy, developed a lightweight elbow exoskeleton prototype, and implemented real-time torque compensation and dynamic load mass prediction using Bi-LSTM and transformer-based multi-task learning models.

Using virtual input rejection to improve control of a platform for characterizing the mechanical properties of human oocytes

J. Abadie, E. Piat

🎯 What it does: Improved the EggSense platform by using Virtual Input Rejection Control (VIRCO) technology to achieve precise control of the position of force-sensing elements in contact with the oocyte.

USVTrack: USV-Based 4D Radar-Camera Tracking Dataset for Autonomous Driving in Inland Waterways

Shanliang Yao, Ryan Wen Liu

Object TrackingAutonomous DrivingMultimodalityPoint CloudBenchmark

🎯 What it does: Created the USVTrack dataset and proposed a pluggable radar-camera matching method RCM;