arXivSub Start free trial

IROS 2025 Papers — Page 18

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

Sound Source Localization for Human-Robot Interaction in Outdoor Environments

Victor Liu, François Grondin

Robotic IntelligenceAudio

🎯 What it does: Adopted a strategy combining microphone arrays with asynchronous close-range microphones to achieve voice source localization in outdoor environments;

Space-Time Graphs of Convex Sets for Multi-Robot Motion Planning

J. Tang, Hang Ma

OptimizationRobotic IntelligenceGraph

🎯 What it does: Proposes a multi-robot motion planning method based on the spatial-temporal convex set graph (ST-GCS), integrating Exact Convex Decomposition (ECD) for trajectory conflict avoidance.

SPADE: Towards Scalable Path Planning Architecture on Actionable Multi-Domain 3D ScenE Graphs

V. Viswanathan, G. Nikolakopoulos

Autonomous DrivingOptimizationGraph Neural NetworkGraph

🎯 What it does: Proposed and implemented the SPADE framework, which utilizes 3D scene graphs to achieve autonomous path planning in dynamic environments. It combines hierarchical planning with local geometric perception, supporting sparse global planning and dense local-level iterative refinement to ensure collision-free movement. By employing directed sampling to eliminate irrelevant edges, the graph planning complexity is reduced. Prioritizing planning at the local level avoids inefficient global replanning caused by path blocking. The effectiveness of the framework in complex dynamic scenarios was verified through simulations and actual deployment on quadruped robots.

SPARK Hand: Scooping-Pinching Adaptive Robotic Hand with Kempe Mechanism for Vertical Passive Grasp in Environmental Constraints

Jiaqi Yin, Wenzeng Zhang

Robotic Intelligence

🎯 What it does: Designed and developed the passive adaptive mechanical fingers SPARK Hand and dual-finger SPARK Hand with two grasping modes: parallel pinching and scooping.

SparseLoc: Sparse Open-Set Landmark-based Global Localization for Autonomous Navigation

Pranjal Paul, K. Krishna

Pose EstimationAutonomous DrivingVision Language ModelSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose the SparseLoc framework for global localization, which leverages a zero-shot vision-language model to generate sparse semantic topological maps and combines Monte Carlo localization with backward optimization to enhance pose estimation.

SparseMeXt: Unlocking the Potential of Sparse Representations for HD Map Construction

Anqing Jiang, Hao Zhao

Autonomous DrivingRepresentation LearningPoint Cloud

🎯 What it does: Systematically review and improve sparse representation techniques in high-precision map construction, proposing a dedicated sparse feature extraction network, a sparse-dense segmentation auxiliary task, and a denoising module based on physical priors.

Spatial Graph Attentional Network Based Place Recognition with Visual Mamba Embedding

Kun Li, Man Qi

RecognitionGraph Neural NetworkTransformerImage

🎯 What it does: Proposed a re-ranking based visual pose recognition framework, including Visual Mamba Embedding (VME) module and Spatial Graph Attention Network (SGAN);

Spatial-Temporal Graph Contrastive Learning with Decreasing Masks for Traffic Flow Forecasting

Bin Ren, Chunhong He

Autonomous DrivingGraph Neural NetworkContrastive LearningTime Series

🎯 What it does: Proposed a masking reduction model called DMSTGCL based on spatial-temporal graph contrastive learning for traffic flow prediction.

Spatio-Temporal Hyperbolic Aggregation Neural Network for Human Action Recognition

Mohamed Sanim Akremi, H. Tabia

RecognitionGraphSequential

🎯 What it does: Propose a new framework that utilizes skeletal data while preserving its tree structure, employing a deep neural network that embeds skeletal joints into hyperbolic space and building a spatiotemporal processing framework on top of it.

Spatiotemporal Motion Prediction of Intraocular Microsurgical Robot in Non-Visible Regions

Yawen Deng, Gui-Bin Bian

Robotic Intelligence

🎯 What it does: Proposed a framework for predicting the tool trajectory of intraocular minimally invasive surgical robots in visual-occluded areas.

Spectral-Temporal Attention for Robust Change Detection

Mayank Thakur, B. N. Patro

TransformerImageTime Series

🎯 What it does: Proposed a spectral-temporal attention network for detecting changes in satellite images, street view images, and indoor environment images across multi-temporal phases, and collected an indoor environment dataset with frequent changes.

Spherical Scissor-Like Reconfigurable Palm Design in Robotic Hands: Insights from Human Hand Functionality

Jiaxi Wang, Zhongxue Gan

OptimizationRobotic Intelligence

🎯 What it does: Based on the spatial reconfigurability of human hands during grasping, a reconfigurable spherical hand was designed, utilizing a spatial scissor mechanism to reshape the hand into various spherical structures using only a single actuator.

SPLATART: Articulated Gaussian Splatting with Estimated Object Structure

Stanley R. Lewis, O. C. Jenkins

Pose EstimationGaussian SplattingImage

🎯 What it does: Proposes the SPLATART workflow, which utilizes Gaussian splat method to learn geometry, color, part segmentation, and joint parameter representations of articulated objects from pose images.

SplatPose: Geometry-Aware 6-DoF Pose Estimation from Single RGB Image via 3D Gaussian Splatting

Linqi Yang, Peng Kang

Pose EstimationGaussian SplattingImage

🎯 What it does: Utilize a single RGB image to achieve 6-DoF pose estimation by combining 3D Gaussian Splatting with a dual-branch neural network.

Splatter Joint: 3D Gaussian Splatting for Articulated Objects

Junyan Li, Wenzhao Lian

GenerationGaussian SplattingImage

🎯 What it does: Proposes the Splatter Joint method, which utilizes 3D Gaussian Splatting to simultaneously model joint motion, appearance, and geometric information from a single-view few images

SPLiCE: Single-Point LiDAR and Camera Calibration & Estimation Leveraging Manhattan World

Minji Kim, Pyojin Kim

Pose EstimationAutonomous DrivingOptimizationImagePoint Cloud

🎯 What it does: Propose an extrinsic calibration method based on single-point LiDAR and a rotatable calibration board, utilizing the Manhattan World (MW) constraint to align LiDAR with the camera.

SplineFormer: An Explainable Transformer Network for Autonomous Endovascular Navigation

Tudor Jianu, Anh Nguyen

Explainability and InterpretabilityRobotic IntelligenceTransformerBiomedical Data

🎯 What it does: Proposed an interpretable Transformer network called SplineFormer, which can represent the continuous shape of a guidewire as B-splines and achieve autonomous vascular navigation under an imitation learning framework.

SpongeBot: A Soft Magnetic Mini-Robot for Controlled Gastric Cell Sampling *

Jiyuan Tian, T. Qiu

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed a soft mini-robot named SpongeBot for active sampling of gastric mucosal cells, integrating sponge structures with magnet actuators for precise control under external wireless magnetic fields.

SR3D: Unleashing Single-view 3D Reconstruction for Transparent and Specular Object Grasping

Mingxu Zhang, Hao Dong

Pose EstimationDepth EstimationRobotic IntelligenceImageMesh

🎯 What it does: Propose a training-free SR3D framework that performs 3D reconstruction using single-view RGB and depth maps, and achieves single-view grasping of transparent and mirror surface objects by localizing objects through view matching and keypoint matching.

SRCNet: Super-resolution Networks for Capsule Endoscope Robots

Menglu Tan, Lin Feng

Super ResolutionRobotic IntelligenceRecurrent Neural NetworkGenerative Adversarial NetworkVideoBiomedical Data

🎯 What it does: Proposed two super-resolution networks, EndoVSR and Bi-RUN, specifically for capsule endoscopy robot videos

Stability Enhancement in Variable Morphing Multi-body AUVs for Underwater Structure Maintenance

Shuai Kang, Longchuan Li

OptimizationRobotic Intelligence

🎯 What it does: Studied a deformable multi-body AUV (VMMAUV) for underwater structure maintenance, utilizing buoyancy regulation and pore angle control mechanisms to optimize the lever center height and enhance system stability.

Stabilizing Humanoid Robot Trajectory Generation via Physics-Informed Learning and Control-Informed Steering

Evelyn D'Elia, Daniele Pucci

Robotic IntelligencePhysics Related

🎯 What it does: An improved imitation learning method is proposed by integrating physical priors and control principles to generate physically constrained smooth humanoid robot trajectories, with a proportional-integral (PI) controller used during inference to reduce drift.

Stable Variable Impedance Control via CLF-MPC for Physical Human-Robot Interaction

Seungmin Choi, Wansoo Kim

OptimizationRobotic Intelligence

🎯 What it does: Proposes a real-time variable damping control parameter prediction method based on CLF-MPC, which ensures system stability in physical human-robot collaboration under parameter variations or external disturbances.

STACKGEN: Generating Stable Structures from Silhouettes via Diffusion

Luzhe Sun, Matthew R. Walter

GenerationRobotic IntelligenceDiffusion modelImage

🎯 What it does: Proposes STACKGEN, a system based on diffusion models for generating stable stacked structures that match a given target outline.

STAGE: A Stream-Centric Generative World Model for Long-Horizon Driving-Scene Simulation

Jiamin Wang, Xinge Zhu

GenerationAutonomous DrivingTransformerWorld ModelVideo

🎯 What it does: Proposed an autoregressive framework called STAGE for long-term driving scene video generation, focusing on solving temporal consistency and feature alignment issues.

Static Analysis and Modeling of a Trunk-Like Robot Capable of Adjustable Multi-Turn Helical Deformation

Zeyu Long, Y. Iwata

OptimizationRobotic Intelligence

🎯 What it does: Analyze the deformation principles of TSSH and establish a static simulation model based on potential energy minimization, subsequently verifying its effectiveness through experiments.

STC-Tracker: Spatiotemporal-Consistent Multi-Robot Collaboration Framework for Long-Term Dynamic Object Tracking

Yanchao Dong, Bin He

Object TrackingRobotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Designed and implemented a multi-robot collaborative dynamic target tracking system called STC-Tracker, which can recover target appearance by replaying historical point clouds from keyframes and monitor target trajectories in real-time.

STEAD: Spatio-Temporal Efficient Anomaly Detection for Time and Compute Sensitive Applications

Andrew Gao, Jun Liu

Anomaly DetectionComputational EfficiencyConvolutional Neural NetworkTransformerVideo

🎯 What it does: Proposes a spatiotemporally efficient anomaly detection method called STEAD, focusing on rapidly detecting anomalies in time- and computation-constrained automated systems

Steady-State Drifting Equilibrium Analysis of Single-Track Two-Wheeled Robots for Controller Design

Feilong Jing, Bin Liang

Robotic Intelligence

🎯 What it does: Extended the drift balance theory to unicycle two-wheeled robots, revealing the mechanism of steady-state drift, and proposed an analytical algorithm based on intrinsic geometry and kinematic relationships. Additionally, a model predictive controller (MPC) was designed based on balance analysis to achieve steady-state drift and balance point transition.

Steering Elongate Multi-legged Robots by Modulating Body Undulation Waves

Esteban Flores, Daniel I. Goldman

Robotic Intelligence

🎯 What it does: Proposed and validated a steering control framework for elongated multi-legged robots based on body swing waveform modulation

STEP Planner: Constructing cross-hierarchical subgoal tree as an embodied long-horizon task planner

Tianxing Zhou, Yufeng Yue

Robotic IntelligenceTransformerLarge Language ModelTextBenchmark

🎯 What it does: Developed the planner STEP with a cross-level subgoal tree to achieve embodied long-term task planning

Stepping Locomotion for a Walking Excavator Robot using Hierarchical Reinforcement Learning and Action Masking*

Ajish Babu, Frank Kirchner

Robotic IntelligenceReinforcement Learning

🎯 What it does: Employ hierarchical reinforcement learning and action masking techniques to train a controller for a walking excavator robot that can navigate over obstacles, steps, and gaps.

STG-Avatar: Animatable Human Avatars via Spacetime Gaussian

Guangan Jiang, Hongyu Wang

GenerationGaussian SplattingOptical FlowVideo

🎯 What it does: Proposed the STG-Avatar framework for reconstructing high-fidelity animatable human avatars from monocular videos, achieving more precise pose control and detail representation by coupling Spacetime Gaussians (STG) with Linear Blend Skinning (LBS), combined with optical flow-driven dynamic region adaptive Gaussian densification.

Stimulating Imagination: Towards General-purpose "Something Something Placement"

Jianyang Wu, Jingmin Chen

Pose EstimationRobotic IntelligenceLarge Language ModelVision Language ModelVision-Language-Action ModelDiffusion modelImageText

🎯 What it does: Propose the SPORT method to achieve general object placement, with the process divided into object localization, target imagination, and robot control.

Stochasticity in Motion: An Information-Theoretic Approach to Trajectory Prediction

Aron Distelzweig, A. Valada

Autonomous Driving

🎯 What it does: Proposes an information-theory-based trajectory prediction method, focusing on uncertainty quantification and decomposition, categorizing uncertainty into aleatoric and epistemic components, and being compatible with existing state-of-the-art motion predictors.

STORM: Spatial-Temporal Iterative Optimization for Reliable Multicopter Trajectory Generation

Jinhao Zhang, Jie Mei

Optimization

🎯 What it does: Proposed a trajectory generation framework for multirotor drones based on spatiotemporal iterative optimization, aiming to improve the safety and computational efficiency of trajectory planning.

Stress-Driven Algorithm for Fiber Alignment in Smart Materials for Controlled Deformation in 4D-Printed Soft Robotics

Won Bin Choi, Wan Kyun Chung

Robotic IntelligenceFibre Orientation DistributionPhysics Related

🎯 What it does: Proposed a path generation strategy that converts external deformation conditions into internal loading conditions for self-activating soft grippers.

Stretchable and High-Precision Optical Tactile Sensor for Trajectory Tracking of Parallel Mechanisms

Yiding Nie, J. S. Dai

Robotic Intelligence

🎯 What it does: Developed a stretchable tactile sensor based on the principle of continuous spectral filtering and integrated it into a planar parallel mechanism to achieve real-time trajectory tracking.

Study on Thunniform Robot Propulsion by Tail-flapping Speed Change

Phan Huy Nam Anh, Su-Hwan Kim

Robotic IntelligencePhysics Related

🎯 What it does: Developed a tuna-inspired robotic fish and proposed a new tail fin beating frequency profile control strategy, dynamically adjusting the tail fin beating speed to enhance propulsion performance.

SubCDM: Collective Decision-Making with a Swarm Subset

Samratul Fuady, M. Soorati

OptimizationRobotic Intelligence

🎯 What it does: Proposed a subset-based collective decision-making method (SubCDM) that uses only a portion of robots for decision-making

Subject-Embedded Vision Transformer with Transfer Learning for Cross-Subject Dynamic Hand Gesture Recognition Using HD-sEMG

Jirou Feng, Jung Kim

RecognitionTransformerBiomedical Data

🎯 What it does: Propose a cross-subject dynamic gesture recognition framework ViT-DHGR based on Vision Transformer, focusing on the gesture transition period to reduce prediction latency, while combining subject embeddings with transfer learning to enhance cross-user generalization.

Subject-Independent sEMG-Based Prosthetic Control Using MAMBA2 with Domain Adaptation

Kihyun Kim, Jiyeon Kang

Domain AdaptationRobotic IntelligenceTransformerBiomedical Data

🎯 What it does: Proposed an adaptive sEMG control method based on the MAMBA2 structure, combining domain adaptation to achieve unlabeled cross-subject wrist movement estimation.

Successor Features for Transfer in Alternating Markov Games

Sunny Amatya, Wenlong Zhang

Reinforcement Learning

🎯 What it does: Explore the knowledge transfer of successor features in zero-sum, perfect information, turn-based Markov games and propose the GGPI algorithm.

SuMag: Suspended Magnetometer Survey for Mineral Data Acquisition with Vertical Take-off and Landing Fixed-wing Aircraft

Robel Efrem, Sajad Saeedi

Physics Related

🎯 What it does: Research on the challenges of installing high-sensitivity magnetometers on drone platforms, proposing solutions and conducting experimental validation to improve data collection in mineral exploration magnetometry.

SuperMag: Vision-based Tactile Data Guided High-resolution Tactile Shape Reconstruction for Magnetic Tactile Sensors

Pei-Lin Hou, Ziyuan Jiao

RestorationSuper ResolutionAuto EncoderImageMultimodality

🎯 What it does: Propose the SuperMag method, which uses high-resolution data from the visual tactile sensor (VBTS) to supervise the super-resolution of the magnetic tactile sensor (MBTS), achieving high-resolution tactile shape reconstruction under low-resolution MBTS inputs.

Surgical D-Knot: Augmented Dexterity for Tying Double Knots by Monitoring Optical Flow in Monocular Attention Windows

Ziyang Chen, K. Goldberg

Robotic IntelligenceOptical FlowImage

🎯 What it does: Developed Surgical D-Knot, using a monocular RGB camera combining learning-based perception and model-based methods to accomplish surgical double-knot tying.

SurgiPose: Estimating Surgical Tool Kinematics from Monocular Video for Surgical Robot Learning

Juo-Tung Chen, Axel Krieger

Pose EstimationRobotic IntelligenceReinforcement LearningVideoBiomedical Data

🎯 What it does: Propose the SurgiPose method, which estimates kinematic information of surgical instruments using monocular video and differentiable rendering, and trains imitation learning strategies based on this information.

Swarm Active Audition with Robots and Drones: Real-World Performance Validation

Kazuhiro Nakadai, Yoko Sasaki

Robotic IntelligenceAudio

🎯 What it does: Proposed and evaluated the collaborative active auditory system SAAS-RD with multiple drones and ground robots, aiming to enhance sound source detection coverage and accessibility in search and rescue operations, with field experiments conducted in real-world environments.

Swept Volume-Based Continuous Object Gathering Trajectory Generation for Tethered Robot Duo

Yuanyuan Du, Shuguang Cui

OptimizationRobotic Intelligence

🎯 What it does: Designed a continuous collection scheme based on sweeping volume, using a double-layer U-shaped modeling for a two-robot team with a rope, and optimizing the trajectory using SVSDF to achieve continuous, collision-safe marine debris collection.

Symmetry-Guided Multi-Agent Inverse Reinforcement Learning

Yongkai Tian, Jie Luo

Reinforcement Learning

🎯 What it does: Developed a generic framework that integrates symmetry into multi-agent adversarial inverse reinforcement learning to improve sample efficiency.

SynthDrive: Scalable Real2Sim2Real Sensor Simulation Pipeline for High-Fidelity Asset Generation and Driving Data Synthesis

Zheng Chen, Qian Zhang

Data SynthesisAutonomous DrivingPrompt EngineeringImageTextMeshRetrieval-Augmented Generation

🎯 What it does: Proposed a scalable Real2Sim2Real sensor simulation pipeline SynthDrive for high-fidelity asset generation and driving data synthesis

Synthetica: Large Scale Synthetic Data Generation for Robot Perception

Ritvik Singh, Ankur Handa

Data SynthesisRobotic IntelligenceTransformerImage

🎯 What it does: Generate 2.7 million synthetic images using a ray tracing renderer to train a real-time, high-precision object detection transformer;

T-CBF: Traversability-based Control Barrier Function to Navigate Vertically Challenging Terrain

Manas Gupta, Xuesu Xiao

Robotic Intelligence

🎯 What it does: Designed and verified a safety navigation method based on Traversability Control Barrier Functions (T-CBF), which uses neural networks to learn safety and unsafe traversability observations for generating safe trajectories, and conducted experiments on simulation and physical Verti-4 Wheeler (V4W) platforms.

T-Touch: a Soft Thermal-haptic Multimodal Fingertip Wearable Device for Immersive Virtual Reality

Youzhan Wang, Chongjing Cao

🎯 What it does: Developed a soft fingertip wearable device called T-Touch, which provides thermal and multi-frequency haptic feedback to enhance VR immersion

TacCap: A Wearable FBG-Based Tactile Sensor for Efficient Human-to-Robot Skill Transfer

Chengyi Xing, M. Cutkosky

Robotic IntelligenceTime Series

🎯 What it does: Developed and evaluated a wearable Fiber Bragg Grating (FBG) tactile sensor called TacCap for achieving seamless transfer between human demonstrations and robotic execution.

TACO: General Acrobatic Flight Control via Target-and-Command-Oriented Reinforcement Learning

Zikang Yin, Shiyu Zhao

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes the TACO framework, which employs goal- and instruction-oriented reinforcement learning to achieve generalizable acrobatic flight control, and enhances the spatiotemporal smoothness and symmetry of the policy through spectral normalization and input-output re-calibration, overcoming the sim-to-real gap.

TACS-Graphs: Traversability-Aware Consistent Scene Graphs for Ground Robot Localization and Mapping

Jeewon Kim, Hyun Myung

Robotic IntelligenceGraph Neural NetworkSimultaneous Localization and MappingGraph

🎯 What it does: Proposed TACS-Graphs, combining ground robot traversability with room segmentation to generate consistent scene graphs.

Tactile sensing soft fingertip with dual air bag structure for an anthropomorphic robotic hand

Jipeng Yin, Yang Yang

Robotic Intelligence

🎯 What it does: Designed a flexible tactile sensing humanoid robot finger with a dual airbag structure, and evaluated its performance in surface texture detection, hard embedding depth detection, object softness/hardness detection, and grasping tasks.

Tactile-based force estimation for interaction control with robot fingers

Elie Chelly, Mahdi Khoramshahi

Robotic Intelligence

🎯 What it does: Propose a data-efficient calibration method to achieve full-array force estimation across different geometries, and use the estimation as online closed-loop control feedback for interaction force tracking in robot fingers.

Tactile-Guided Robotic Ultrasound: Mapping Preplanned Scan Paths for Intercostal Imaging

Yifan Zhang, Zhongliang Jiang

Robotic IntelligencePoint CloudBiomedical DataComputed TomographyUltrasound

🎯 What it does: Developed a tactile-guided robotic ultrasound system utilizing tactile point cloud generation and interpolation, registration with image-based dense bone surface point clouds, automatic tilt angle adjustment, and other techniques to achieve precise scanning path mapping of pulmonary fissures.

TagGuideBot: Enhancing Robot Intelligence with Object Tags and VLMs

Jiayi Chen, F. Yu

Robotic IntelligencePrompt EngineeringVision Language Model

🎯 What it does: Designed a framework named Tag-GuideBot, which integrates visual language models (VLM) and object tags with location point prompts and robot motion planning models to more accurately understand and execute complex commands, thereby improving the efficiency and naturalness of human-robot interaction.

TAR: Teacher-Aligned Representations via Contrastive Learning for Quadrupedal Locomotion

Amr Mousa, Richard Allmendinger

Robotic IntelligenceReinforcement LearningContrastive Learning

🎯 What it does: Proposed a contrastive learning-based teacher-aligned representation framework TAR to improve sample efficiency and generalization ability for quadruped robot locomotion based on reinforcement learning.

Target Handling Modalities with Obstacle Avoidance for Planar Soft Growing Manipulator Design

Ozan Nurcan, Fabio Stroppa

OptimizationRobotic Intelligence

🎯 What it does: Building upon the existing planar soft-growing manipulator design optimizer, this study introduces five new target processing modes tailored for real-world operational scenarios. By treating targets as spatial obstacles, the approach enables effective planning and design for diverse operational tasks.

Target Localization and Following Based on LiDAR and Ultra-Wideband Ranging with Consideration of Target Visibility

Lin Guo, Chau Yuen

Object TrackingAutonomous DrivingOptimizationSimultaneous Localization and MappingPoint CloudSequential

🎯 What it does: Proposed a sequence matching method based on LiDAR and UWB ranging for target localization, and integrated a visibility objective function into the Dynamic Window Approach (DWA) to achieve visibility-aware tracking path planning;

TartanGround: A Large-Scale Dataset for Ground Robot Perception and Navigation

Manthan Patel, Wenshan Wang

Robotic IntelligenceSimultaneous Localization and MappingImageMultimodalityPoint Cloud

🎯 What it does: Proposes the TartanGround large-scale multimodal dataset to enhance ground robots' perception and navigation in diverse environments.

TASeg: Text-aware RGB-T Semantic Segmentation based on Fine-tuning Vision Foundation Models

Meng Yu, Yufeng Yue

SegmentationTransformerSupervised Fine-TuningVision Language ModelImageTextMultimodality

🎯 What it does: Proposes the TASeg framework, achieving text-aware RGB-T semantic segmentation through LoRA fine-tuning of visual foundation models; introduces a dynamic feature fusion module (DFFM) in the image encoder to integrate multi-modal visual features while freezing the original Transformer blocks of SAM; utilizes CLIP-generated text embeddings in the mask decoder for semantic alignment, correcting classification errors and enhancing semantic understanding accuracy.

Task-Aware Robotic Grasping by evaluating Quality Diversity Solutions through Foundation Models

Aurel X. Appius, Stéphane Doncieux

OptimizationRepresentation LearningRobotic IntelligenceTransformerLarge Language ModelImageBenchmark

🎯 What it does: This paper proposes a zero-shot task-aware grasping framework that combines large language models (LLMs) with quality diversity (QD) algorithms, leveraging semantic and geometric dual representations to achieve task-conditioned grasping, achieving a 73.6% task-conditioned grasping region prediction IoU on the YCB dataset.

Task-driven SLAM Benchmarking for Robot Navigation

Yanwei Du, Patricio A. Vela

Robotic IntelligenceSimultaneous Localization and MappingBenchmark

🎯 What it does: Proposed and implemented TaskSLAM-Bench, a task-based SLAM benchmarking platform focusing on localization accuracy and mapping capabilities, evaluating modern visual and LiDAR SLAM methods in both simulated and real environments.

Task-Guided and Object-Centric Conditioning for Effective and Adaptive Diffusion Policy

Wenshuo Wang, Haiyue Zhu

Representation LearningRobotic IntelligenceTransformerDiffusion modelImage

🎯 What it does: Propose the TOC-DP framework, which combines SlotAttention with task-specific segmentation priors to achieve object-based visual representation learning, and refines the representations with action-aware processing during downstream policy learning.

Task-Oriented Token Pruning for Efficient Object Detection and Segmentation

Hao Liang, Xilin Chen

Object DetectionSegmentationComputational EfficiencyTransformerImagePoint Cloud

🎯 What it does: Proposed a task-oriented Token pruning method called TaskTP, which can dynamically adjust the retention ratio of Tokens based on the target category set to improve computational efficiency in detection and segmentation tasks.

TBAP: Tapping-Based Auditory Perception for Identifying Container Materials

Zehao Li, Wenbo Ding

ClassificationAudio

🎯 What it does: A framework for container material identification combining deep learning with active auditory perception is proposed, along with a modular robotic system capable of achieving three-dimensional full-scale tapping.

TCNet: A Temporally Consistent Network for Self-supervised Monocular Depth Estimation

Ying Zhu, Mengyuan Liu

Depth EstimationImage

🎯 What it does: Propose TCNet, which leverages temporal consistency to improve monocular self-supervised depth estimation.

Teacher Motion Priors: Enhancing Robot Locomotion over Challenging Terrain

Fangcheng Jin, Zhengtao Zhang

Knowledge DistillationRobotic IntelligenceGenerative Adversarial Network

🎯 What it does: Propose a teacher-prior framework based on the teacher-student paradigm, enhancing robotic mobility on complex terrains through imitation learning and auxiliary task learning;

Team Orienteering Problem with Communication Constraints

Marco Túlio P. T. Tristão, D. Macharet

Optimization

🎯 What it does: Proposes a multi-objective team cruising problem model that balances task coverage, communication quality, and energy consumption while optimizing under a fixed budget.

Tele-GS: 3D Gaussian Scene Representation for Low-Bandwidth Teleoperation

Chunyang Zhao, Danwei Wang

Autonomous DrivingComputational EfficiencyGaussian SplattingPoint Cloud

🎯 What it does: Proposed a LiDAR-fusion based 3D Gaussian Splatting (3DGS) model for remote operations in low-bandwidth environments; first construct a static point cloud map as initial Gaussians through LiDAR semantic mapping, then during remote operations only render the pre-built 3DGS and transmit safety-critical information such as vehicle pose and dynamic objects in real time, significantly reducing data transmission requirements while maintaining a realistic remote experience.

Teleoperated Teaching of Task and Impedance (TTTI): Multi-Modal Interface Extending Haptic Device for Robotic Skill Transfer

Astrid W.E. Rots, Luka Peternel

Robotic IntelligenceMultimodality

🎯 What it does: A multimodal interface was developed to enable remote teaching of low-level impedance regulation skills and high-level task decision-making skills through a handheld haptic device, validated on a shelf-stacking robot in a simulated supermarket environment.

TEM3-Learning: Time-Efficient Multimodal Multi-Task Learning for Advanced Assistive Driving

Wenzhuo Liu, Wenshuo Wang

ClassificationRecognitionAutonomous DrivingMultimodality

🎯 What it does: Propose the TEM3-Learning framework, which jointly optimizes driver emotion recognition, driver behavior recognition, traffic context recognition, and vehicle behavior recognition through a two-phase architecture.

Temporal-Spatial Representation Fusion for Dexterous Manipulation Learning with Unpaired Visual-Action Data

Guwen Han, Qi Ye

Robotic IntelligenceRecurrent Neural NetworkReinforcement LearningAuto Encoder

🎯 What it does: Propose the UnVALe framework, which learns human dexterous manipulation skills using unpaired action data, and employs LSTM and VAE to reconstruct temporal and spatial action priors, using them as rewards to guide reinforcement learning (RL).

Tendon-driven Grasper Design for Aerial Robot Perching on Tree Branches

Haichuan Li, B. B. Kocer

Object DetectionRobotic Intelligence

🎯 What it does: Proposed a bio-inspired aerial platform that uses a claw-like tendon-driven mechanism to perch on branches, minimizing energy consumption during data collection; employs a raptor-inspired visual algorithm to locate tree trunks and horizontal branches; achieves safe attachment between the platform and branches through passive compliant tendon mechanisms.

TERL: Large-Scale Multi-Target Encirclement Using Transformer-Enhanced Reinforcement Learning

Heng Zhang, Xiaoqiang Ren

Robotic IntelligenceTransformerReinforcement Learning

🎯 What it does: Proposes a Transformer-Enhanced Reinforcement Learning (TERL) framework for large-scale multi-target encirclement problems.

TerraX: Visual Terrain Classification Enhanced by Vision-Language Models

Hongze Li, Huijing Zhao

ClassificationVision Language ModelContrastive LearningMultimodalityBenchmark

🎯 What it does: Propose the TerraX framework, integrating vision-language models for visual terrain classification, constructing the TerraData dataset, TerraBench evaluation benchmark, and implementing the TerraCLIP model.

TetraGrip: Sensor-Driven Multi-Suction Reactive Object Manipulation in Cluttered Scenes

Paolo Torrado, Joshua R. Smith

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Developed a TetraGrip multi-suction gripper system with four suction cups, equipped with linear actuators and optical ToF sensors, and evaluated its performance in a warehouse-like environment.

TeX-NeRF: Neural Radiance Fields for Novel HADAR View Synthesis

Chonghao Zhong, Hao Zhao

GenerationPose EstimationNeural Radiance FieldImage

🎯 What it does: Proposed TeX-NeRF, a NeRF framework based on thermal images for novel view HADAR view synthesis.

TextInPlace: Indoor Visual Place Recognition in Repetitive Structures with Scene Text Spotting and Verification

Huaqi Tao, Hong Zhang

RetrievalConvolutional Neural NetworkImageTextBenchmark

🎯 What it does: Proposes the TextInPlace framework for indoor visual place recognition, addressing visual ambiguity in repetitive structures by combining scene text detection and verification.

TFRR: A Novel Tensegrity-Based Fracture Reduction Robot with Force Sensing

Chenguang Cui, Fanny Ficuciello

Robotic Intelligence

🎯 What it does: Proposed a fracture reduction robot based on tensegrity structure, integrated with force sensing and control functions.

That’s Iconic! Designing Augmented Reality Iconic Gestures To Enhance Multi-modal Communication For Morphologically Limited Robots

Yifei Zhu, Tom Williams

Robotic IntelligenceMultimodality

🎯 What it does: Conducted a human subject study in an open environment to compare the differences in task performance and subjective perception among four AR-assisted symbolic gestures (humanoid, non-humanoid, pointing symbolic, abstract).

The Anti-Misalignment Mechanism of Bionic Knee Joint of Lower Limb Exoskeleton Based on Spherical Cross Four-Bar

Jiaxun Wu, Hongyuan Zhang

OptimizationRobotic Intelligence

🎯 What it does: Proposed and optimized a biomimetic knee joint structure based on a spherical cross four-bar linkage to reduce misalignment between the exoskeleton and the human knee joint, and enhance usability comfort and safety.

The Art of Not Getting Smacked: ISO/TS 15066-Compliant Variable Admittance Control for Safe Human-Robot Interaction

Matteo Nini, F. Ferraguti

Safty and PrivacyRobotic Intelligence

🎯 What it does: Proposes a new framework to achieve safe and reliable human-robot physical interaction through online adaptive variable damping controller parameters, meeting the ISO/TS 15066 standard.

The Common Objects Underwater (COU) Dataset for Robust Underwater Object Detection

Rishi Mukherjee, Junaed Sattar

Object DetectionImageBenchmark

🎯 What it does: Created and made publicly available the COU (Common Objects Underwater) dataset, containing approximately 10K instance-segmented underwater man-made object images, covering diverse environments such as closed water pools and open lakes and seas.

The Constitutional Filter: Bayesian Estimation of Compliant Agents

Simon Kohaut, K. Kersting

Object TrackingTime SeriesSequential

🎯 What it does: Proposed a Bayesian estimation method based on the neuro-symbolic model Constitution called Constitutional Filter (CoFi), for improving agent tracking.

The Duke Humanoid: Design and Control for Energy-Efficient Bipedal Locomotion Using Passive Dynamics

Boxi Xia, Boyuan Chen

Robotic IntelligenceReinforcement Learning

🎯 What it does: Designed and implemented Duke Humanoid—a 10-degree-of-freedom (DOF), open-source humanoid robot platform—and developed zero-shot deployment reinforcement learning (RL) strategies for speed-tracking walking tasks; additionally, proposed an end-to-end RL algorithm that encourages the robot to leverage passive dynamics to improve energy efficiency.

The Foundation for Tactile Robots: Approaching the Holistic Analysis of a Robot’s Force Sensing Capabilities

R. Kirschner, Sami Haddadin

Robotic IntelligenceTabular

🎯 What it does: Studied the effects of different joint configurations, temperature, and other factors on the force perception performance of a 7-degree-of-freedom robot, and introduced a force perception error mapping tool for structured research and comparison of robotic force perception capabilities.

The Impact of Autonomy Levels and System Errors on Cognitive Load and Trust in Human-Robot Collaborative Tasks

Juan Jose Garcia Cardenas, Adriana Tapus

Robotic IntelligenceTabularTime Series

🎯 What it does: Participants were asked to place four bottles of different shapes into a box within three minutes using a UR5 robotic arm under three control conditions. Physiological indicators (blink rate, GSR, facial temperature) and task performance (success rate, completion time) were measured to evaluate the impact of different levels of autonomy and system errors on cognitive load and trust.

The Impact of VR and 2D Interfaces on Human Feedback in Preference-Based Robot Learning

Jorge de Heuvel, Maren Bennewitz

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningText

🎯 What it does: Systematically compared the impact of VR and 2D interfaces in robot learning, collecting 2,325 preference data.

The KIT Robotic Hands – A Scalable Humanoid Hand Platform With Multi-Modal Sensing and In-Hand Embedded Processing

J. Starke, Tamim Asfour

Robotic IntelligenceImageMultimodality

🎯 What it does: Designed and implemented an expandable KIT robotic hand platform with multimodal sensing, hand-embedded processing, adaptive lower drive mechanism, and continuously controllable thumb rotation to enhance dexterity, and validated its scalability through two application examples: ARMAR-7 and ARMAR-DE.

The Monado SLAM Dataset for Egocentric Visual-Inertial Tracking

Mateo de Mayo, Taihú Pire

Simultaneous Localization and MappingVideoBenchmark

🎯 What it does: Provides a real-world sequence dataset called Monado SLAM tailored for challenging scenarios in visual inertial odometry and SLAM with head-mounted sensors.

The Parallel Pneumatic Artificial Muscle Platform Based on RBF Neural Network Compensation

Jun Li, Yinhui Xie

Robotic Intelligence

🎯 What it does: This study proposes a control scheme combining a radial basis function (RBF) neural network with an adaptive learning rate and a PID controller, based on a two-degree-of-freedom parallel mechanism with four pneumatic muscles. The scheme automatically outputs X and Y axis pressure values according to desired angles, enabling precise driving of the joint to reach specified angles.

The Sampling-Gaussian for Stereo Matching

Baiyu Pan, Jun Cheng

Depth EstimationImage

🎯 What it does: Propose Sampling-Gaussian as an alternative to soft-argmax to achieve accuracy improvement without additional inference time

THE-SEAN: A Heart Rate Variation-Inspired Temporally High-Order Event-Based Visual Odometry with Self-Supervised Spiking Event Accumulation Networks

Chaoran Xiong, Ling Pei

Pose EstimationAutonomous DrivingSpiking Neural NetworkSimultaneous Localization and MappingSequential

🎯 What it does: Proposed a temporal high-order event-driven visual odometry called THE-SEAN based on self-supervised spiking neural networks, which can dynamically adjust the estimation trigger decisions according to motion and environmental changes.

Thermal Updraft Profiling with an Array of Show Drones

Pedro Lacerda, Máté Nagy

Physics Related

🎯 What it does: Utilize a multi-rotor drone array for distributed measurement of atmospheric thermal updrafts to estimate updraft velocity.