arXivSub Start free trial

ICRA 2025 Papers — Page 16

IEEE International Conference on Robotics and Automation · 1604 papers

Tunable Virtual IMU Frame by Weighted Averaging of Multiple Non-Collocated IMUs

Yizhou Gao, Tim D. Barfoot

OptimizationTime Series

🎯 What it does: Proposed a method to fuse multiple physically separated IMUs into a single virtual IMU (VIMU) through weighted averaging.

TWIN: Two-handed Intelligent Benchmark for Bimanual Manipulation

Markus Grotz, Dieter Fox

Data SynthesisRobotic IntelligenceBenchmark

🎯 What it does: Proposed a bimanual manipulation benchmark (TWIN) capable of automatically generating training data, containing 13 new tasks and 23 task variations, with open-source code;

UA-PnP: Uncertainty-Aware End-to-End Bird's Eye View Visual Perception and Prediction for Autonomous Driving

Zijian Huang, Qi Hao

Autonomous DrivingTransformer

🎯 What it does: Propose an uncertainty-aware, end-to-end visual perception and prediction framework based on bird's eye view (BEV), comprising a feature distribution estimation network, an uncertainty-based Transformer, and a generator for future instance segmentation with uncertainty evidence decoder.

UAD: Unsupervised Affordance Distillation for Generalization in Robotic Manipulation

Yihe Tang, Fei-Fei Li

Knowledge DistillationRobotic IntelligenceLarge Language ModelVision Language ModelImageText

🎯 What it does: Proposes an unsupervised usability distillation method (UAD) that leverages large visual models and vision-language models to automatically annotate data, trains only a lightweight task-conditioned decoder, and uses the generated usability as the observation space to train imitation learning policies, demonstrating generalization capabilities in real-world scenarios and human activities.

UASTHN: Uncertainty-Aware Deep Homography Estimation for UAV Satellite-Thermal Geo-Localization

Jiuhong Xiao, Giuseppe Loianno

Pose EstimationConvolutional Neural NetworkImage

🎯 What it does: Proposed a framework named UASTHN for depth homography estimation in drone satellite thermal imaging geolocation, which employs CropTTA (Crop-based Test Time Augmentation) to measure data uncertainty and combines Deep Ensembles to evaluate model uncertainty.

UAV-Assisted Self-Supervised Terrain Awareness for Off-Road Navigation

Jean-Michel Fortin, Philippe Giguère

Autonomous DrivingImage

🎯 What it does: Utilizing hovering drones to acquire aerial perspective images for self-supervised learning, training ground vehicle vibration, roughness, and energy consumption prediction models, and collecting a 2.8 km ground and aerial image dataset in a forest environment, verifying the method's practical application in path planning.

UDSV: Unsupervised Deep Stitching for Tractor-Trailer Surround View

Leyao Sun, Mengyin Fu

Autonomous DrivingConvolutional Neural NetworkImage

🎯 What it does: Proposed an unsupervised deep stitching method for the tractor-trailer surround view system, and designed a feature extraction module (FMT) and a spatiotemporal consistent control point constraint strategy (STCC) tailored for low-overlap scenarios.

UGotMe: An Embodied System for Affective Human-Robot Interaction

Peizhen Li, Runze Yang

Computational EfficiencyRobotic IntelligenceImageMultimodality

🎯 What it does: Designed and implemented a embodied emotional human-computer interaction system named UGotMe, specialized for multi-party dialogue scenarios. It enhances real-time responsiveness through two denoising strategies (extracting speakers' facial images from raw images and employing a customized active facial extraction strategy to filter non-active speakers) and efficient data transmission, with deployment verification on the Ameca humanoid robot.

ULOC: Learning to Localize in Complex Large-Scale Environments with Ultra-Wideband Ranges

Thien-Minh Nguyen, Lihua Xie

Autonomous DrivingTransformerSimultaneous Localization and Mapping

🎯 What it does: Proposed the ULOC learning framework for positioning in large-scale, complex environments using UWB. First, deploy anchor points with unknown positions, collect UWB observations during vehicle movement, and use VIO, LIO, and other SLAM methods to align with a prior map, generating map-consistent pose estimates as training labels. Subsequently, a MAMBA-based network learns UWB ranging patterns.

UltraFastCrackSeg: A Lightweight Real-Time Crack Segmentation Model with Task-Oriented Pretraining

Weiqing Qi, Yang Yang

SegmentationComputational EfficiencyConvolutional Neural NetworkImage

🎯 What it does: Proposed the UltraFastCrackSeg lightweight real-time crack segmentation model

Ultrasound-Guided Robotic Blood Drawing and In Vivo Studies on Submillimetre Vessels of Rats

Shuaiqi Jing, Peng Qi

Robotic IntelligenceBiomedical DataUltrasound

🎯 What it does: Proposed an ultrasound-guided robotic blood collection system capable of performing vascular puncture in fine blood vessels with a diameter of approximately 0.7 mm in a rat's tail.

UncAD: Towards Safe End-to-end Autonomous Driving via Online Map Uncertainty

Pengxuan Yang, Dongbin Zhao

Autonomous DrivingSimultaneous Localization and MappingMultimodalityPoint Cloud

🎯 What it does: Developed the UncAD system to estimate online map uncertainty and use it for motion prediction and planning, proposing a trajectory selection strategy based on uncertainty-aware collision perception.

Uncertainty-Aware Deep Reinforcement Learning with Calibrated Quantile Regression and Evidential Learning

A. Stutts, A. Trivedi

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Propose a novel statistical method that introduces uncertainty-awareness in model-agnostic distributed deep reinforcement learning, applicable to task and safety-critical robots.

Uncertainty-aware Probabilistic 3D Human Motion Forecasting via Invertible Networks

Yue Ma, Xiaohui Liang

Pose EstimationFlow-based ModelTime SeriesSequential

🎯 What it does: Propose the ProbHMI model, using reversible networks to achieve 3D human motion prediction and explicitly estimate uncertainty.

Uncertainty-Aware Probabilistic Risk Quantification of SOTIF for Autonomous Vehicles

Botao Yao, Shaoming Duan

Autonomous Driving

🎯 What it does: Proposed an uncertainty-aware probabilistic risk assessment framework for safety-of-the-intended-functionality (SOTIF) in autonomous vehicles (AVs), to quantify the risk of AVs violating safety constraints and calculate the expected average severity of such violations.

Uncertainty-Aware Visual-Inertial SLAM with Volumetric Occupancy Mapping

Jaehyung Jung, Stefan Leutenegger

Pose EstimationDepth EstimationAutonomous DrivingSimultaneous Localization and MappingImageMultimodality

🎯 What it does: Propose a visual-inertial SLAM method that integrates sparse reprojection error, IMU preintegration, relative pose factors, and dense volumetric occupancy mapping

Uncertainty-Guided Enhancement on Driving Perception System Via Foundation Models

Yunhao Yang, Ben Snyder

Autonomous DrivingComputational Efficiency

🎯 What it does: By quantifying the predictive uncertainty of driving perception models, the method intelligently invokes a multi-modal foundation model only when the uncertainty exceeds a threshold to refine predictions, and incorporates a time reasoning mechanism to further improve accuracy.

Under Pressure: Altimeter-Aided ICP for 3D Maps Consistency

William Dubois, François Pomerleau

OptimizationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a new method that integrates barometer height constraints into the ICP algorithm to reduce drift in restricted environments such as vertical shafts.

Understanding Dynamic Human-Robot Proxemics in the Case of Four-Legged Canine-Inspired Robots

Xiangmin Xu, Robin Bretin

Pose EstimationRobotic Intelligence

🎯 What it does: Studying the close-range behaviors of quadruped robots toward human participants under different interaction postures using a motion capture system, and analyzing human-robot distance in dynamic scenes

Underwater Motions Analysis and Control of a Coupling-Tiltable Unmanned Aerial-Aquatic Vehicle

Dongyue Huang, Ben M. Chen

OptimizationRobotic Intelligence

🎯 What it does: Analyzed the underwater motion characteristics of the self-developed UAAV Mirs-Alioth and designed a corresponding controller, with experimental validation of its effectiveness.

UniAff: A Unified Representation of Affordances for Tool Usage and Articulation with Vision-Language Models

Qiaojun Yu, Cewu Lu

Representation LearningRobotic IntelligenceLarge Language Model

🎯 What it does: Proposed the UniAff framework, integrating 3D object-centered manipulation with task understanding, and constructed an annotated dataset containing 900 articulated objects and 600 tools; used multi-modal large language models (MLLMs) to infer object-centered representations for grasping and motion constraint reasoning.

UniBEVFusion: Unified Radar-Vision Bevfusion for 3D Object Detection

Haocheng Zhao, Yutao Yue

Object DetectionAutonomous DrivingGaussian SplattingMultimodality

🎯 What it does: Propose the UniBEVFusion network, integrating the Radar Depth Lift-Splat-Shoot (RDL) module, Unified Feature Fusion (UFF) method, and Failure Test (FT) failure test experiment for radar-visual fusion in 3D object detection.

Unified Adaptive and Cooperative Planning Using Multi-Task Coregionalized Gaussian Processes

Lorenzo Booth, Stefano Carpin

OptimizationRobotic Intelligence

🎯 What it does: Proposes a multi-robot information path planner that combines environmental kernels and task kernels, utilizing multi-output Gaussian processes to unify spatial-temporal environmental priors with observation correlations between sensing vehicles, in order to obtain samples that maximize map accuracy improvement.

Unified Human Localization and Trajectory Prediction with Monocular Vision

Po-Chien Luan, Alexandre Alahi

Pose EstimationAutonomous DrivingTransformerSimultaneous Localization and MappingImage

🎯 What it does: Propose MonoTransmotion (MT), a Transformer-based framework that utilizes a monocular camera to simultaneously achieve human BEV localization and trajectory prediction.

Universal Online Temporal Calibration for Optimization-Based Visual-Inertial Navigation Systems

Yunfei Fan, Feng Zhou

Autonomous DrivingOptimizationSimultaneous Localization and MappingVideoMultimodalityTime Series

🎯 What it does: Proposed a general online time calibration strategy that incorporates time offset as a state parameter into an optimized residual model to achieve time synchronization in visual inertial navigation systems.

Unleashing HyDRa: Hybrid Fusion, Depth Consistency and Radar for Unified 3D Perception

Philipp Wolters, Gerhard Rigoll

Depth EstimationAutonomous DrivingTransformerImagePoint CloudBenchmark

🎯 What it does: Proposes the HyDRa camera-radar fusion architecture for low-cost, vision-oriented 3D perception tasks.

Unlock the Power of Unlabeled Data in Language Driving Model

Chaoqun Wang, Ruimao Zhang

Autonomous DrivingTransformerPrompt EngineeringVision Language ModelMultimodalityBenchmark

🎯 What it does: Propose a method that utilizes template prompts and self-consistency refinement to generate pseudo answers from unlabeled data, and constructs a language driving model using a pre-trained VisionLLM.

Unlocking Potential: Gaze-Based Interfaces in Assistive Robotics for Users with Severe Speech and Motor Impairment

Himanshu Vishwakarma, Pradipta Biswas

Robotic IntelligenceOptical FlowVideo

🎯 What it does: Developed a gaze-control-based robotic system to assist users with severe speech and motor impairments in completing stamp printing tasks.

Unsupervised Domain Adaptation for Gait State Estimation

R. L. Medrano, Elliott J. Rouse

Pose EstimationDomain AdaptationTime SeriesBiomedical Data

🎯 What it does: Propose an unsupervised learning method to estimate the gait states of exoskeleton wearers based on kinematic measurements without requiring labeled data, and address the domain adaptation problem from regular gait to exoskeleton-assisted gait.

Unveiling the Black Box: Independent Functional Module Evaluation for Bird's-Eye-View Perception Model

Ludan Zhang, Keqiang Li

Autonomous DrivingExplainability and InterpretabilityAuto Encoder

🎯 What it does: Proposed a framework named BEV-IFME, which quantifies the similarity between feature maps of functional modules and ground truth in a unified semantic representation space to evaluate the training maturity of each functional module.

Unveiling the Depths: A Multi-Modal Fusion Framework for Challenging Scenarios

Jialei Xu, Xiangyang Ji

Depth EstimationMultimodality

🎯 What it does: Propose a depth estimation framework based on multi-modal fusion, which independently calculates rough depth maps using RGB and long-wave infrared images, then generates confidence maps for potential depth regions through a self-supervised confidence prediction network, and finally achieves robust depth estimation by end-to-end fusion of depth information from both modalities.

Updating Robot Safety Representations Online From Natural Language Feedback

L. Santos, Andrea V. Bajcsy

Robotic IntelligenceReinforcement Learning from Human FeedbackVision Language ModelImageTextMultimodality

🎯 What it does: Use a vision-language model to interpret natural language feedback and robot image observations, online update the robot's representation of safety constraints, and update the Hamilton-Jacobi reachability safety controller through efficient hot-starting technology.

UPTor: Unified 3D Human Pose Dynamics and Trajectory Prediction for Human-Robot Interaction

Nisarga Nilavadi, Timm Linder

Pose EstimationRobotic IntelligenceGraph Neural NetworkTransformerTime SeriesSequential

🎯 What it does: Propose a unified method that leverages short-sequence input poses to simultaneously predict the dynamics and motion trajectories of human keypoints.

UpViTaL: Unpaired Visual-Tactile Self-Supervised Representation Learning for Dexterous Robotic Manipulation

Guwen Han, Qi Ye

Representation LearningRobotic IntelligenceReinforcement LearningMultimodalityTime Series

🎯 What it does: Proposes the UpViTaL method, which utilizes unpaired visual and tactile self-supervised representation learning, combining a time-series tactile module with visual pretraining, and fuses the two representations through an RL reward mechanism to achieve dexterous manipulation of a robotic hand without tactile sensors;

UR-MVO: Robust Monocular Visual Odometry for Underwater Scenarios

Zein Alabedeen Barhoum, S. Kolyubin

Convolutional Neural NetworkTransformerSupervised Fine-TuningSimultaneous Localization and MappingImage

🎯 What it does: Proposed a robust monocular visual odometry (UR-MVO) pipeline for underwater scenarios, using SuperPoint for feature extraction and SuperGlue for matching, while employing few-shot unsupervised learning to fine-tune SuperPoint for scene-specific adaptation. Additionally, a semantic segmentation model trained on underwater images is integrated to remove unreliable features from dynamic objects and the background.

Use the Force, Bot! - Force-Aware ProDMP with Event-Based Replanning

Paul Werner Lödige, Rudolf Lioutikov

Robotic Intelligence

🎯 What it does: Proposes FA-ProDMP, which can adjust trajectories in real-time during operation based on measured and desired forces, for contact-rich manipulation tasks.

User-Aware Collaborative Learning in Human-Robot Interactions

Balint Gucsi, Long Tran-Thanh

Robotic IntelligenceReinforcement LearningMultimodality

🎯 What it does: Studied a user-oriented collaborative learning framework aimed at enhancing human-machine collaboration experience by minimizing human frustration.

Using Physiological Measures, Gaze, and Facial Expressions to Model Human Trust in a Robot Partner

Haley N. Green, Tariq Iqbal

ClassificationRobotic IntelligenceMultimodalityBiomedical Data

🎯 What it does: Designed and implemented a human-robot supervision interaction experiment, collecting physiological and visual data such as peripheral capillary oxygen saturation, galvanic skin response, skin temperature, gaze, and facial expressions, and utilized machine learning models to identify objective indicators that best represent trust in the robot partner.

V-Pilot: A Velocity Vector Control Agent for Fixed-Wing UAVs from Imperfect Demonstrations

Xudong Gong, Huaimin Wang

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed the V-Pilot agent, which achieves velocity vector control for fixed-wing unmanned aerial vehicles using reinforcement learning under imperfect demonstration conditions;

V2X-DG: Domain Generalization for Vehicle-to-Everything Cooperative Perception

Baolu Li, Hongkai Yu

Domain AdaptationAutonomous DrivingPoint Cloud

🎯 What it does: Study the domain generalization problem of LiDAR-based V2X collaborative perception, propose CMAG and CFC methods, and verify them on four public datasets.

V2X-DGW: Domain Generalization for Multi-Agent Perception Under Adverse Weather Conditions

Baolu Li, Hongkai Yu

Object DetectionDomain AdaptationAutonomous DrivingContrastive LearningPoint Cloud

🎯 What it does: Propose a V2X-DGW method based on domain generalization for LiDAR multi-agent 3D object detection under adverse weather conditions while maintaining high performance in clear weather

VAIR: Visuo-Acoustic Implicit Representations for Low-Cost, Multi-Modal Transparent Surface Reconstruction in Indoor Scenes

A. Sethuraman, Katherine A. Skinner

ImageMultimodalityPoint CloudMeshUltrasoundAudio

🎯 What it does: Constructing a low-cost multi-modal transparent surface reconstruction method by fusing acoustic and visual modalities through implicit neural representations, achieving dense reconstruction of indoor transparent surfaces.

Valg: Vision-Based Adaptive Laser Gripper for Model-Free Pose Control of Floating Objects at Air-Liquid Interface

Xusheng Hui, Haonan You

Pose EstimationRobotic IntelligenceOptical FlowImagePhysics Related

🎯 What it does: Proposed a vision-based adaptive laser gripper VALG for non-contact control of the attitude of floating objects at the air-liquid interface.

Validity Learning on Failures: Mitigating the Distribution Shift in Autonomous Vehicle Planning

Fazel Arasteh, Kasra Rezaee

Autonomous DrivingBenchmark

🎯 What it does: Designed a method called Validity Learning on Failures (VL(on failure)), which learns to identify valid trajectories in the current environment by generating failure scenarios with a pre-trained planner and constructing a failure dataset.

Variable-Friction In-Hand Manipulation for Arbitrary Objects via Diffusion-Based Imitation Learning

Qiyang Yan, Adam J. Spiers

Robotic IntelligenceReinforcement LearningDiffusion model

🎯 What it does: Propose a diffusion-based imitation learning method to achieve autonomous manipulation of objects with arbitrary shapes to any target pose on variable friction hands.

Variable-Stiffness Nasotracheal Intubation Robot with Passive Buffering: A Modular Platform in Mannequin Studies

Ruoyi Hao, Hongliang Ren

Robotic Intelligence

🎯 What it does: A modular, variable stiffness nasotracheal intubation robot was developed, and its feasibility was validated through phantom experiments.

VascularPilot3D: Toward a 3D Fully Autonomous Navigation for Endovascular Robotics

Jingwei Song, Maani Ghaffari

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed the VascularPilot3D endovascular robot full 3D autonomous navigation system

Vegetable Peeling: A Case Study in Constrained Dexterous Manipulation

Tao Chen, Pulkit Agrawal

Robotic IntelligenceAgriculture Related

🎯 What it does: Proposed a constrained dexterous manipulation method for food peeling, learning a redirecting controller that supports subsequent peeling tasks.

Verifiably Following Complex Robot Instructions with Foundation Models

Benedict Quartey, G. Konidaris

Robotic IntelligenceLarge Language ModelText

🎯 What it does: Propose the LIMP method, enabling robots to verifiably execute complex, open-ended instructions in real-world environments without pre-built semantic maps.

Versatile Distributed Maneuvering With Generalized Formations Using Guiding Vector Fields

Yang Lu, Xin Xu

Robotic Intelligence

🎯 What it does: Propose a unified approach to achieve multi-functional maneuvering of distributed robots in general formation patterns by decomposing robot maneuvers into independent interception and encirclement components. The method derives a singularity-free guidance vector field (GVF) and a distributed coordination mechanism based on consensus theory, using two virtual coordinates as dimensions of an abstract manifold. This guides robots to perform formation tracking, target encirclement, orbiting, and other motions, with a controller designed for non-fully actuated robot models based on GVF.

VertiCoder: Self-Supervised Kinodynamic Representation Learning on Vertically Challenging Terrain

Mohammad Nazeri, Xuesu Xiao

Representation LearningRobotic IntelligenceTransformerAuto Encoder

🎯 What it does: Proposes VertiCoder, a self-supervised representation learning method for vertical challenging terrain robot mobility, capable of simultaneously accomplishing four downstream tasks with a single representation: forward dynamics learning, inverse dynamics learning, behavior cloning, and patch reconstruction.

Vibrotactile Haptics with Soft Magnetoresponsive Surface Interface

Evan Rimer, Matthew Robertson

🎯 What it does: Explore the use of magnetoresponsive silicone to generate tactile vibrations and develop the VibroFlex Pad for experimental testing and user evaluation

VideoSAM: Open-World Video Segmentation

Pinxue Guo, Wenqiang Zhang

SegmentationVideo

🎯 What it does: Proposed the VideoSAM framework to address object cross-frame association and segmentation fine-grained inconsistency in open-world video segmentation.

VIP-Dock: Vision, Inertia, and Pressure Sensor Fusion for Underwater Docking with Optical Beacon Guidance

Suohang Zhang, Yanhu Chen

OptimizationRobotic IntelligenceImageMultimodalityPhysics Related

🎯 What it does: Propose the VIP-Dock optical beacon tracking algorithm, integrating visual, inertial, and pressure sensors to achieve real-time tracking for underwater AUV docking.

VisFly: An Efficient and Versatile Simulator for Training Vision-Based Flight

Fanxing Li, Danping Zou

Autonomous DrivingComputational EfficiencyReinforcement Learning

🎯 What it does: Developed VisFly, a highly efficient simulator for training vision-based quadrotor flight control strategies.

Vision Transformers for End-to-End Vision-Based Quadrotor Obstacle Avoidance

Anish Bhattacharya, Vijay Kumar

Autonomous DrivingComputational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkTransformerImage

🎯 What it does: Proposed and demonstrated an end-to-end high-speed visual obstacle avoidance method based on an attention mechanism.

Vision-Based Fuzzy Control System with Intention Detection for Smart Walkers: Enhancing Usability for Stroke Survivors with Unilateral Upper Limb Impairments

Mahdi Chalaki, Mahdi Tavakoli

Pose EstimationRobotic IntelligenceImage

🎯 What it does: A smart cane using a fuzzy control algorithm was developed, which identifies the turning intention of users with unilateral limb dysfunction through shoulder abduction angles, and is equipped with force sensors and stereo cameras to enhance responsiveness.

Vision-Based Movement Primitives for Lunar Hazard Avoidance

Joseph M. Cloud, Jason M. Schuler

Robotic IntelligenceImage

🎯 What it does: Using Dynamic Motion Primitives (DMP) to represent the driving mode of the ISRU test excavator, and integrating a real-time vision-based obstacle avoidance system to perform hazard avoidance on the lunar surface.

Visual Lidar Recursive Online Tracker (ViLiROT) for Autonomous Surface Vessels

Henrik Hilmarsen, Rudolf Mester

Object TrackingAutonomous DrivingOptical FlowImagePoint Cloud

🎯 What it does: Proposed a multi-sensor fusion pipeline based on LiDAR and camera data for multi-target tracking in autonomous surface vessels.

Visual-Based Forklift Learning System Enabling Zero-Shot Sim2Real Without Real-World Data

Koshi Oishi, Seigo Ito

Domain AdaptationRobotic IntelligenceReinforcement LearningImage

🎯 What it does: Proposed and implemented a vision-based deep reinforcement learning system, training forklifts in a digital simulation environment, and subsequently verifying zero-shot Sim2Real performance on a 1/14 scale real robot forklift.

Visual-Linguistic Reasoning for Pedestrian Trajectory Prediction

Dereje Shenkut, B.V.K. Vijaya Kumar

Autonomous DrivingRecurrent Neural NetworkTransformerSupervised Fine-TuningPrompt EngineeringVideo

🎯 What it does: Fine-tune pre-trained vision-language models (VLM) on prompts related to road scenes and pedestrians to learn semantic scene context and high-level reasoning features; subsequently combine these VLM features with the pedestrian's past trajectory history through an encoder-decoder structure to predict future trajectories.

Visually Robust Adversarial Imitation Learning from Videos with Contrastive Learning

Vittorio Giammarino, I. Paschalidis

Contrastive LearningVideo

🎯 What it does: Proposed the C-LAIfO algorithm to address visual discrepancy issues in video imitation learning.

Visuo-Tactile Object Pose Estimation for a Multi-Finger Robot Hand With Low-Resolution in-Hand Tactile Sensing

Lukas Mack, Joerg Stueckler

Pose EstimationOptimizationRobotic IntelligenceImageMultimodality

🎯 What it does: Proposes a visuo-tactile object pose estimation method for a multi-fingered robotic arm that integrates vision, proprioception, and low-resolution binary tactile sensing, optimized through probabilistic modeling in a factor graph format;

ViTa-Zero: Zero-shot Visuotactile Object 6D Pose Estimation

Hongyu Li, T. Padır

Pose EstimationRobotic IntelligenceMultimodality

🎯 What it does: Proposed the ViTa-Zero zero-shot visual tactile 6D pose estimation framework

VITaL Pretraining: Visuo-Tactile Pretraining for Tactile and Non-Tactile Manipulation Policies

Abraham George, A. Farimani

Representation LearningRobotic IntelligenceMultimodality

🎯 What it does: Pre-train the imitation learning platform with visual-tactile inputs to enhance performance in grasping and manipulation tasks.

ViViDex: Learning Vision-Based Dexterous Manipulation from Human Videos

Zerui Chen, I. Laptev

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningDiffusion modelVideoPoint Cloud

🎯 What it does: Learning a unified vision-based control strategy for multi-fingered robotic hands to operate various objects under multiple poses

VizFlyt: Perception-centric Pedagogical Framework For Autonomous Aerial Robots

Kushagra Srivastava, N. Sanket

Robotic IntelligenceGaussian SplattingImage

🎯 What it does: Developed and validated the VizFlyt testing framework, achieving real-time realistic visual sensor simulation through external positioning system attitude and 3D Gaussian Splatting technology, supporting safe autonomous drone control algorithm testing, and proposing an open-course system based on this framework.

VLM-GroNav: Robot Navigation Using Physically Grounded Vision-Language Models in Outdoor Environments

Mohamed Bashir Elnoor, Dinesh Manocha

Robotic IntelligenceVision Language Model

🎯 What it does: Proposes an outdoor autonomous robot navigation algorithm called VLM-GroNav that can handle various terrain traversability conditions.

VLM-Vac: Enhancing Smart Vacuums Through VLM Knowledge Distillation and Language-Guided Experience Replay

Reihaneh Mirjalili, Wolfram Burgard

Knowledge DistillationRobotic IntelligenceVision Language Model

🎯 What it does: Designed a framework called VLM-Vac, which enhances the autonomy of smart vacuum cleaners through zero-shot object detection and knowledge distillation of Vision-Language Models (VLM), and implements continuous learning via language-guided experience replay.

VLN-KHVR: Knowledge-And-History Aware Visual Representation for Continuous Vision-and-Language Navigation

Ping Kong, Zhibo Pang

Autonomous DrivingRepresentation LearningVision-Language-Action ModelImageTextMultimodalityRetrieval-Augmented Generation

🎯 What it does: Propose a knowledge and history-aware visual representation method called VLN-KHVR for continuous vision-language navigation.

vMF-Contact: Uncertainty-Aware Evidential Learning for Probabilistic Contact-Grasp in Noisy Clutter

Yitian Shi, R. Rayyes

Robotic Intelligence

🎯 What it does: Proposes a noise-free evidence learning method called vMF-Contact for 6-DoF grasping, capable of simultaneously capturing aleatoric uncertainty from data noise and epistemic uncertainty in out-of-distribution (OOD) recognition;

VQA-Driven Event Maps for Assistive Navigation for People with Low Vision in Urban Environments

Joseph Morales, Jordi Sanchez-Riera

Autonomous DrivingComputational EfficiencyPrompt EngineeringImage

🎯 What it does: Proposed a VQA-based event map-assisted navigation framework for low vision individuals in urban environments.

VSB - Variable Stiffness Based on Bowden Cables: A Simple Mechanism for Soft Robotic Hands

Steffen Puhlmann, Hannes Höppner

Robotic Intelligence

🎯 What it does: Proposed and verified a simple variable stiffness robotic hand design based on Bowden cables, utilizing tension-based antagonistic actuation to connect the elastic component with the servo motor.

VSS-SLAM: Voxelized Surfel Splatting for Geometally Accurate SLAM

Xuanhua Chen, Xingshuo Wang

Gaussian SplattingSimultaneous Localization and MappingImage

🎯 What it does: Propose a VSS-SLAM method based on voxelized surfels for incremental mapping in unknown environments.

Wallbounce: Push Wall to Navigate with Contact-Implicit MPC

Xiaohan Liu (Carnegie Mellon University), Ralph Hollis (Carnegie Mellon University)

Robotic Intelligence

🎯 What it does: Proposes a framework that utilizes non-periodic contact to achieve high maneuverability locomotion.

WaLTER: A Wheel and Leg Tumbling Expedition Robot

David Jay, Jonathan E. Clark

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed and implemented a hybrid wheg-legged quadruped robot named WaLTER, developed an intuitive remote control scheme, and validated its control method using deep reinforcement learning (DRL). The MuJoCo multi-body simulation platform and a 2.1 kg physical prototype were constructed to conduct experiments on rough terrain navigation and energy consumption.

Watch Less, Feel More: Sim-to-Real RL for Generalizable Articulated Object Manipulation via Motion Adaptation and Impedance Control

Tan-Dzung Do, He Wang

Domain AdaptationRobotic IntelligenceReinforcement Learning

🎯 What it does: This paper proposes a pipeline based on reinforcement learning, combining variable damping control and motion adaptation, leveraging observation history to achieve general manipulation of articulated objects, and achieving smooth and precise actions in zero-shot sim-to-real transfer.

Watch Your STEPP: Semantic Traversability Estimation Using Pose Projected Features

Sebastian Ægidius, Dimitrios Kanoulas

Anomaly DetectionRobotic IntelligenceTransformerAuto EncoderImage

🎯 What it does: Leverages human walking demonstrations, combined with pixel-level features generated by the DINOv2 vision Transformer, to estimate the traversability of legged robots on different terrains and perform anomaly detection through an encoder-decoder MLP and reconstruction loss.

Wavelet-Based Distributed Coverage for Heterogeneous Agents

Ananya Rao, David Wettergreen

OptimizationRobotic Intelligence

🎯 What it does: Proposed an algorithm using waveform transformation for heterogeneous agent distributed coverage based on isentropic metric to balance exploration and exploitation

Wcdt: World-Centric Diffusion Transformer for Traffic Scene Generation

Chen Yang, Arsalan Heydarian

GenerationData SynthesisAutonomous DrivingTransformerDiffusion modelMultimodality

🎯 What it does: Proposes a World-Centered Diffusion Transformer (WcDT) framework for autonomous driving trajectory generation, combining diffusion probability models and transformers to optimize the entire generation process.

Weakly-Supervised Learning via Multi-Lateral Decoder Branching for Tool Segmentation in Robot-Assisted Cardiovascular Catheterization

O. Omisore, Lei Wang

SegmentationRobotic IntelligenceConvolutional Neural NetworkBiomedical Data

🎯 What it does: A weakly supervised learning method based on multi-side pseudo labels is developed for cardiovascular vascular imaging tool segmentation.

Wearable Soft Sensing Band with Stretchable Sensors for Torque Estimation and Hand Gesture Recognition

Junhwan Choi, Jung Kim

ClassificationRecognitionTime SeriesBiomedical Data

🎯 What it does: Designed and evaluated a wearable flexible stretchable sensor band for estimating muscle activity through changes in muscle volume, including isometric force estimation and gesture recognition.

Weathergs: 3D Scene Reconstruction in Adverse Weather Conditions Via Gaussian Splatting

Chenghao Qian, Gustav Markkula

RestorationGaussian SplattingImage

🎯 What it does: Proposed the WeatherGS framework, based on 3D Gaussian Splatting, which reconstructs clear 3D scenes under adverse weather by preprocessing multi-view images.

Whenever, Wherever: Towards Orchestrating Crowd Simulations with Spatio-Temporal Spawn Dynamics

Thomas Kreutz, Alejandro Sánchez Guinea

Data Synthesis

🎯 What it does: Proposed the nTPP-GMM model, combining Neural Temporal Point Processes with spawn-conditional Gaussian Mixture Model to simulate the spatiotemporal dynamics of crowd generation, achieving more realistic crowd simulation.

Whisker-Based Active Tactile Perception for Contour Reconstruction

Yixuan Dang, A. Knoll

Robotic Intelligence

🎯 What it does: Developed an active contour reconstruction method based on hair-like tactile sensors, including the magnetic sensing construction of the sensor, gradient descent extraction and Bayesian filtering of the tip contact position, as well as an active motion control strategy to maintain the optimal relative posture of the sensor with respect to the object surface.

Whole-Body Control Through Narrow Gaps from Pixels to Action

Tianyue Wu, Fei Gao

OptimizationRobotic IntelligenceReinforcement LearningImage

🎯 What it does: Explored a pure data-driven control method for multirotors to fly through narrow gaps using pixel and body-aware perception.

Whole-Body End-Effector Pose Tracking

Tifanny Portela, Marco Hutter

Pose EstimationRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a whole-body reinforcement learning approach to achieve end-effector pose tracking on rough irregular terrain with a large workspace.

Whole-Body Model Predictive Control for Mobile Manipulation With Task Priority Transition

Yushi Wang, Mingguo Zhao

OptimizationRobotic Intelligence

🎯 What it does: Proposed a Whole-Body Model Predictive Control (MPC) framework for control in mobile manipulation tasks involving different time scales, integrating task priorities across task and time dimensions to achieve seamless priority switching.

WildFusion: Multimodal Implicit 3D Reconstructions in the Wild

Yanbaihui Liu, Boyuan Chen

Representation LearningNeural Radiance FieldImageMultimodalityPoint CloudAudio

🎯 What it does: Propose the WildFusion method, which utilizes multimodal implicit neural representations for 3D scene reconstruction in wild environments.

WildLMa: Long Horizon Loco-Manipulation in the Wild

Ri-Zhao Qiu, Xiaolong Wang

Robotic IntelligenceLarge Language ModelVision-Language-Action ModelContrastive LearningImageText

🎯 What it does: Propose the WildLMa framework, integrating low-level controller adaptation, generalizable visual-motor skill libraries, and LLM planning interfaces to achieve long-term autonomous navigation and manipulation in dynamic environments.

Winding Number-Guided Edge-Preserving Implicit Neural Representation of CAD Surfaces

Yuhang Cheng, Xiaogang Wang

Neural Radiance FieldMesh

🎯 What it does: Proposed a ring number guided implicit surface reconstruction method that combines a ring number guided regularizer and a dynamic edge sampling strategy;

Word2Wave: Language Driven Mission Programming for Efficient Subsea Deployments of Marine Robots

Ruo Chen, Md Jahidul Islam

Robotic IntelligenceTransformerLarge Language ModelPrompt EngineeringText

🎯 What it does: Developed a language-based interface called Word2Wave for dynamic task programming and parameter configuration in autonomous underwater vehicles (AUVs), enabling interactive programming for remote underwater tasks.

Work Smarter Not Harder: Simple Imitation Learning with CS-PIBT Outperforms Large-Scale Imitation Learning for MAPF

Rishi Veerapaneni, Maxim Likhachev

Optimization

🎯 What it does: A simple machine learning MAPF strategy was trained using the CS-PIBT post-processing technique and compared with existing ML MAPF strategies.

World Model-Based Perception for Visual Legged Locomotion

Hang Lai, Weinan Zhang

Robotic IntelligenceWorld ModelMultimodality

🎯 What it does: Proposed and implemented a visual legged gait control method based on a world model (WMP), which trains the world model in simulation and uses it to predict real-world trajectories to guide the policy.

Would You Trust Me Now? A Study on Trust Repair Strategies in Human-Robot Collaboration

Joséphine Mélot-Chesnel, M. D. Graaf

Robotic Intelligence

🎯 What it does: Conduct experimental evaluation of the effectiveness of three trust repair strategies (apology, denial, compensation) under two categories of trust breach (competence breach and integrity breach)

X-MOBILITY: End-to-End Generalizable Navigation via World Modeling

Wei Liu, Soha Pouya

Autonomous DrivingWorld Model

🎯 What it does: Propose X-Mobility, an end-to-end generalizable navigation model that achieves effective training through world modeling, a rich multi-head decoder, and decoupled action strategies.

XMoP: Whole-Body Control Policy for Zero-Shot Cross-Embodiment Neural Motion Planning

P. Rath, N. Gopalan

Robotic Intelligence

🎯 What it does: Propose the XMoP neural policy, which learns to plan actions across a distribution of robot variants and achieves zero-shot transfer to unseen robots

You Only Estimate Once: Unified, One-stage, Real-Time Category-Level Articulated Object 6D Pose Estimation for Robotic Grasping

Jingshun Huang, X. Xue

Pose EstimationRobotic IntelligencePoint Cloud

🎯 What it does: Studied 6D pose estimation for category-level deformable objects, proposing the YOEO single-stage end-to-end method, which can simultaneously output instance segmentation and NPCS representations.

ZeroBP: Learning Position-Aware Correspondence for Zero-Shot 6D Pose Estimation in Bin-Picking

Jianqiu Chen, Zhenyu He

Pose EstimationRobotic Intelligence

🎯 What it does: To address the zero-shot 6D pose estimation problem in bin picking tasks, the ZeroBP framework is proposed for pose inference.

ZeroCAP: Zero-Shot Multi-Robot Context Aware Pattern Formation via Large Language Models

Vishnunandan L. N. Venkatesh, Byung-Cheol Min

SegmentationRobotic IntelligenceTransformerLarge Language ModelText

🎯 What it does: Developed the ZeroCAP system, integrating large language models with multi-robot systems to achieve zero-shot context-aware pattern generation.