arXivSub Start free trial

IROS 2025 Papers — Page 14

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

Offline Imitation Learning upon Arbitrary Demonstrations by Pre-Training Dynamics Representations

Haitong Ma, Na Li

Representation LearningReinforcement LearningContrastive Learning

🎯 What it does: Propose introducing a pre-training phase in offline imitation learning to enhance performance when expert data is limited by learning dynamic representations.

Offline motion tracking of multi-link mechanisms using inertial sensor fusion and EKF-preconditioned FGO

Aderajew Tilahun, G. Dissanayake

OptimizationRobotic IntelligenceTime Series

🎯 What it does: This paper proposes an offline spatial motion estimation method for multi-link mechanisms based on IMU sensor fusion.

Offline Reinforcement Learning with Koopman Operators for Control of Soft Robots

Yue Jiang, Xinglong Zhang

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a soft robot control framework based on the Koopman operator and offline reinforcement learning, capable of generating control policies without requiring a physical simulator or real-world interaction;

Omni-Scan: Creating Visually-Accurate Digital Twin Object Models Using a Bimanual Robot with Handover and Gaussian Splat Merging

Tianshuang Qiu, Kenneth Y. Goldberg

Anomaly DetectionRobotic IntelligenceGaussian SplattingOptical FlowImage

🎯 What it does: Proposed the Omni-Scan pipeline, utilizing dual-robotic hands for grasping and handover, combined with a single fixed camera to rotate and re-grasp objects, achieving 360° multi-view image acquisition without occlusion, and generating high-quality 3D Gaussian Splatting (3DGS) digital twin models through model training.

OmniPose6D: Towards Short-Term Object Pose Tracking in Dynamic Scenes from Monocular RGB

Yunzhi Lin, Kevin J. Liang

Object TrackingPose EstimationImageBenchmark

🎯 What it does: Proposed a large synthetic dataset Omni-Pose6D, a benchmark evaluation framework, and designed a keypoint refinement network that utilizes uncertainty modeling.

On Learning Closed-Loop Probabilistic Multi-Agent Simulator

Juanwu Lu, Ziran Wang

Autonomous DrivingWorld ModelImagePoint Cloud

🎯 What it does: Proposed a probabilistic multi-agent simulation framework based on a hierarchical Bayesian model, Neural Interactive Agents (NIVA), achieving closed-loop, observation-conditioned autoregressive sampling;

On learning racing policies with reinforcement learning

Grzegorz Czechmanowski, Krzysztof Walas

Autonomous DrivingReinforcement Learning

🎯 What it does: Proposes a method using reinforcement learning to learn autonomous racing strategies

On the Analysis of Stability, Sensitivity and Transparency in Variable Admittance Control for pHRI Enhanced by Virtual Fixtures

Davide Tebaldi, L. Biagiotti

Robotic Intelligence

🎯 What it does: Analyze the instability sources of proxy-type restricted impedance controllers in physical human-robot interaction, conduct sensitivity analysis to identify the impact of control parameters on overall system stability, propose adaptive proxy parameter techniques to maximize transparency, and validate the results through simulation and experiments.

On the Design of Fast-Response Variable-Stiffness Continuum Robot with Electro-permanent Magnet-based Ball Joints

Taerim Lee, Donghyun Hwang

Robotic Intelligence

🎯 What it does: Designed a fast-response variable stiffness continuum robot based on electromagnetic permanent magnets (EPM) spherical joints, and verified its performance through experiments.

On the Design, Analysis, and Experimental Validation of Pneumatically-Actuated, Soft Robotic, Telescopic Structures

Bryan Busby, Minas V. Liarokapis

Robotic Intelligence

🎯 What it does: Studied the design parameters of soft mechanical expandable structures and their effects on expansion, speed, and smoothness, and experimentally validated these relationships.

On the Role of Jacobians in Robust Manipulation

Joshua T. Grace, A. Dollar

Robotic Intelligence

🎯 What it does: Proposed a new inverse Jacobian estimation method and a precise manipulation control method without prior knowledge, verified on the Yale Model O hand, Yale Stewart hand, and UR5e arm.

On the Vulnerability of LLM/VLM-Controlled Robotics

Xiyang Wu, A. S. Bedi

Robotic IntelligenceLarge Language ModelVision Language Model

🎯 What it does: Investigated the vulnerability of robot systems controlled by large language models (LLMs) and vision-language models (VLMs) under minor variations in input modalities, and proposed corresponding mathematical models and experimental validation methods.

On-Board Vision-Language Models (VLMs) for Personalized Motion Control of Autonomous Vehicles

Can Cui, Ziran Wang

Autonomous DrivingVision Language ModelRetrieval-Augmented Generation

🎯 What it does: Propose a lightweight in-vehicle vision-language model framework to achieve low-latency personalized driving control.

On-Chip Dynamic Mechanical Characterization: from Cells to Nucleus*

Jingjin Ge, Xiaoming Liu

VideoBiomedical Data

🎯 What it does: Propose a chip-based dynamic mechanical characterization method using microchannels, achieving automated measurement of mechanical properties of cancer and normal cells through high-frequency imaging and computer vision, revealing that nuclear mechanical properties determine overall cellular mechanical behavior.

On-Demand Motion Conversion of Magnetic Helical Microrobots Using Chemistry- and Microstructural-Modified Surface Wettability Modulation

Yaozhen Hou, Huaping Wang

Robotic IntelligencePhysics Related

🎯 What it does: Achieving on-demand motion conversion of magnetic helical microrobots in different liquid environments through surface chemistry and microstructure modification.

One-Shot Affordance Grounding of Deformable Objects in Egocentric Organizing Scenes

Wanjun Jia, Zhiyong Li

RecognitionObject DetectionMeta LearningPrompt EngineeringImage

🎯 What it does: Propose a one-shot deformable object localization method (OS-AGDO) that identifies unknown deformable objects in egocentric scenes with minimal samples.

One-Shot Gesture Recognition for Underwater Diver-To-Robot Communication*

Rishikesh Joshi, Junaed Sattar

RecognitionRobotic IntelligenceImageVideo

🎯 What it does: Proposed a single-demonstration-based underwater diver-robot mirror gesture recognition method.

One-shot Global Localization through Semantic Distribution Feature Retrieval and Semantic Topological Histogram Registration

Feixuan Huang, Heng Zhao

Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposed a single-shot global localization method based on LiDAR semantic maps

One-Shot Robust Imitation Learning for Long-Horizon Visuomotor Tasks from Unsegmented Demonstrations

Shaokang Wu, Yanlong Huang

Robotic IntelligenceMeta LearningVideo

🎯 What it does: Proposes the MiLa framework, which utilizes dynamic motion primitives and meta-learning to achieve segmentation-free learning for long-horizon visual motion tasks, enabling rapid adaptation to unseen tasks through a single demonstration; simultaneously, the method demonstrates resilience to external disturbances and visual occlusions.

Online 6DoF Global Localisation in Forests using Semantically-Guided Re-Localisation and Cross-View Factor-Graph Optimisation

Lucas Carvalho de Lima, Milad Ramezani

Pose EstimationConvolutional Neural NetworkSimultaneous Localization and MappingImage

🎯 What it does: Propose the FGLoc6D method to achieve 6DoF global localization and online pose estimation for ground robots in forest environments with weakened GPS, utilizing deep semantic-guided relocalization and cross-view factor graph optimization.

Online Anti-Swing Trajectory Refinement for Variable-Length Cable-Suspended Aerial Transportation Robot

Haixin Yu, Xiao Liang

OptimizationRobotic Intelligence

🎯 What it does: Proposed an online trajectory refinement method for a drone transportation robot with variable-length cable suspension, achieving real-time reference trajectory optimization to suppress payload swinging.

Online Concurrent Multi-Robot Coverage Path Planning

Ratijit Mitra, Indranil Saha

OptimizationRobotic IntelligenceBenchmark

🎯 What it does: Proposed a centralized online multi-robot coverage path planning algorithm not based on a field-of-view window, supporting simultaneous planning and execution

Online Estimation of Table-Top Grown Strawberry Mass in Field Conditions with Occlusions

Jinshan Zhen, Ya Xiong

RestorationSegmentationDepth EstimationConvolutional Neural NetworkGenerative Adversarial NetworkImagePoint CloudAgriculture Related

🎯 What it does: Proposes a visual pipeline based on RGB-D and deep learning for online real-time estimation of desktop-planted strawberry mass, combining instance segmentation, occlusion completion, and tilt angle correction, then mapping geometric features to mass using polynomial regression.

Online Hierarchical Policy Learning using Physics Priors for Robot Navigation in Unknown Environments

Wei Chen, A. H. Qureshi

Robotic IntelligenceReinforcement LearningPhysics Related

🎯 What it does: Propose a hierarchical policy learning based on physical priors for robot navigation in unknown indoor environments.

Online Imitation Learning for Manipulation via Decaying Relative Correction through Teleoperation

Cheng Pan, Josie Hughes

Robotic IntelligenceReinforcement LearningAgriculture Related

🎯 What it does: A 6-degree-of-freedom (6-DOF) spatial correction system based on cable-driven remote operation is proposed, along with the introduction of a Decaying Relative Correction (DRC) method. This method temporarily applies spatial offset vectors provided by experts to policy trajectories, reducing the number of expert interventions and improving the success rate of robotic manipulators in operational tasks such as strawberry picking and cloth wiping through an online imitation learning framework.

Online Iterative Learning with Forward Simulation for Sub-minimum End-effector Displacement Positioning

Weiming Qu, Dingsheng Luo

Robotic Intelligence

🎯 What it does: Propose a novel high-precision robotic arm operation framework based on online iterative learning and forward simulation, capable of achieving sub-minimal end-effector displacement positioning.

Online Motion Planning for Quadrotor Multi-Point Navigation Using Efficient Imitation Learning-Based Strategy

Jin Zhou, Shuo Li

OptimizationRobotic IntelligenceSequential

🎯 What it does: Propose an online scheme based on imitation learning to enable quadrotor drones to achieve near real-time optimal trajectory navigation among multiple target points.

Online Residual Model Learning for Model Predictive Control of Autonomous Surface Vehicles in Real-World Environments

Arturo Gamboa-Gonzalez, Wei Wang

Autonomous DrivingOptimization

🎯 What it does: Proposed an online residual model learning framework that integrates offline simulation learning with real-time online learning into nonlinear model predictive control (MPC) for autonomous surface vehicle control.

Online Synthesis of Control Barrier Functions with Local Occupancy Grid Maps for Safe Navigation in Unknown Environments

Yuepeng Zhang, Xiang Yin

Autonomous Driving

🎯 What it does: In unknown environments, real-time online synthesis of control barrier functions (CBF) from local occupancy grid maps (OGM).

Online-HMM with Two-Layer Bayesian Method for Operator's Expected Speed Estimation in Teleoperated Gluing Tasks *

Wenke Zhou, Yu Wang

Computational EfficiencyRobotic Intelligence

🎯 What it does: Developed an online HMM combined with a two-layer Bayesian method to suppress operator's physiological tremors and estimate their desired speed.

OpAC: An Optimization-Augmented Control Framework for Single and Coordinated Multi-Arm Robotic Manipulation

Melih Özcan, Ozgur S. Oguz

OptimizationRobotic Intelligence

🎯 What it does: Proposed a multi-modal control framework OpAC, combining force control with optimization-enhanced motion planning to sequentially complete single-arm and multi-arm robotic manipulation tasks.

Open-loop Deep Reinforcement Learning Control of Soft Robotic In-hand Manipulations

Gabriel Suske, Oliver Sawodny

Robotic IntelligenceReinforcement Learning

🎯 What it does: Designed and implemented a novel anthropomorphic soft gripper, utilizing deep reinforcement learning to synthesize an open-loop controller, achieving the movement and manipulation of objects within the hand.

Open-Set LiDAR Panoptic Segmentation Guided by Uncertainty-Aware Learning

Rohit Mohan, Abhinav Valada

SegmentationAutonomous DrivingContrastive LearningPoint Cloud

🎯 What it does: Proposes a uncertainty-guided open-set LiDAR panoptic segmentation framework called ULOPS, which models prediction uncertainty using Dirichlet evidence learning and identifies and segments unknown instances during inference through uncertainty.

Open-World Task Planning for Humanoid Bimanual Dexterous Manipulation via Vision-Language Models

Zixin Tang, Fei Chen

Robotic IntelligenceVision Language ModelMultimodalityBenchmark

🎯 What it does: Proposes a large-scale benchmark called OBiMan-Bench for evaluating open-world task planning in dual-arm dexterous manipulation, and designs a zero-shot planning framework called OBiMan-Planner based on vision-language models, which includes two modules: scenario normalization and task planning;

OpenFusion++: An Open-vocabulary Real-time Scene Understanding System

Xiaofeng Jin, Matteo Matteucci

Object DetectionSegmentationDepth EstimationPoint CloudMesh

🎯 What it does: Proposes OpenFusion++, a real-time 3D semantic geometric reconstruction system based on TSDF for open-vocabulary scene understanding.

OpenGS-Fusion: Open-Vocabulary Dense Mapping with Hybrid 3D Gaussian Splatting for Refined Object-Level Understanding

Dianyi Yang, Yi Yang

SegmentationVision Language ModelGaussian Splatting

🎯 What it does: Proposed OpenGS-Fusion, an open-vocabulary dense mapping framework that enhances semantic modeling and refines 3D object-level understanding.

OpenMIGS: Multi-granularity Information-preserving Open-Vocabulary 3D Gaussian Splatting

Jingyu Zhao, Yufeng Yue

GenerationRetrievalGaussian SplattingPoint Cloud

🎯 What it does: Proposes the OpenMIGS framework, achieving multi-granularity, information-preserving open-vocabulary 3D Gaussian rendering through constructing object-level Gaussian fields and lightweight implicit fields.

OpenNav: Open-World Navigation with Multimodal Large Language Models

Mingfeng Yuan, Steven L. Waslander

Autonomous DrivingRobotic IntelligenceTransformerLarge Language ModelVision Language ModelVision-Language-Action ModelMultimodality

🎯 What it does: Propose a zero-shot visual-language navigation framework that leverages a multi-modal large language model (MLLM) to parse free-text instructions and generate top-down 2D value maps to synthesize trajectory points, enabling diverse navigation tasks with open-set instructions and objects.

OpenObject-NAV: Open-Vocabulary Object-Oriented Navigation Based on Dynamic Carrier-Relationship Scene Graph

Yujie Tang, Yufeng Yue

Autonomous DrivingRobotic IntelligenceGraph Neural NetworkLarge Language ModelReinforcement LearningVision-Language-Action Model

🎯 What it does: Propose an object navigation method based on open-vocabulary Carrier-Relationship Scene Graph (CRSG), and validate its effectiveness in the Habitat simulator and on real robots.

OpenRoboCare: A Multimodal Multi-Task Expert Demonstration Dataset for Robot Caregiving

Xiaoyu Liang, T. Bhattacharjee

Pose EstimationRobotic IntelligenceMultimodalityBenchmark

🎯 What it does: Collected the multimodal dataset OpenRoboCare, recording expert demonstrations by 21 occupational therapists performing 15 daily living activities on two mannequins.

OpenVox: Real-time Instance-level Open-vocabulary Probabilistic Voxel Representation

Yinan Deng, Yufeng Yue

Object DetectionSegmentationComputational EfficiencyRepresentation LearningVision Language ModelSimultaneous Localization and MappingImageText

🎯 What it does: Proposes OpenVox, a real-time incremental open-vocabulary probabilistic instance voxel representation, with a frontend employing efficient instance segmentation and a language reasoning pipeline based on title encoding, and a backend implementing probabilistic instance voxels and cross-frame incremental fusion, including instance association and real-time map evolution.

Opportunistic Collaborative Planning with Large Vision Model Guided Control and Joint Query-Service Optimization

Jiayi Chen, Chengzhong Xu

Autonomous DrivingOptimization

🎯 What it does: Proposes an opportunistic collaborative planning (OCP) method that integrates an efficient local model with a cloud-based large visual model to enable autonomous vehicle navigation in open scenarios.

Opt-in Camera: Person Identification in Video via UWB Localization and Its Application to Opt-in Systems

Matthew Ishige, Ryo Yonetani

RecognitionSafty and PrivacySimultaneous Localization and MappingVideo

🎯 What it does: Developed an opt-in camera system that utilizes UWB tag tracking and visual identification to record only individuals who have explicitly consented to being recorded

Opti-Acoustic Scene Reconstruction in Highly Turbid Underwater Environments

Ivana Collado-Gonzalez, Brendan Englot

Convolutional Neural NetworkImageMultimodalityAudio

🎯 What it does: A real-time photoacoustic scene reconstruction method is proposed, specifically optimized for turbid water environments, utilizing image region of interest (ROI) matching with sonar data, and combining sonar distance and camera height information for reconstruction.

OptiGrasp: Optimized Grasp Pose Detection Using RGB Images for Warehouse Picking Robots

Simranjeet Singh, Joshua R. Smith

Pose EstimationRobotic IntelligenceConvolutional Neural NetworkSupervised Fine-TuningImage

🎯 What it does: Achieve suction grasping pose detection using only RGB images and foundation models, achieving an 81.9% success rate on a real robot

Optimal Control of Walkers with Parallel Actuation

Ludovic De Matteïs, Nicolas Mansard

OptimizationRobotic Intelligence

🎯 What it does: Achieving more precise generation of walking motion by explicitly considering the kinematic constraints of the closed-loop chain in optimal control problems.

Optimal Motion Scaling for Delayed Telesurgery

Jason Lim, Michael C. Yip

OptimizationRobotic Intelligence

🎯 What it does: Conduct user experiments to investigate the relationship between network latency, motion scaling factors, and teleoperation surgery performance, and propose a personalized scaling factor mapping model

Optimal Scheduling of a Dual-Arm Robot for Efficient Strawberry Harvesting in Plant Factories

Yuankai Zhu, Chen Peng

OptimizationAgriculture Related

🎯 What it does: Propose a Mixed Integer Linear Programming (MILP) framework for systematic scheduling and coordination of dual-arm harvesting robots, minimizing the total harvesting completion time by utilizing pre-mapped fruit positions.

Optimal Trajectory Planning in a Vertically Undulating Snake Locomotion using Contact-implicit Optimization

Adarsh Salagame, Alireza Ramezani

OptimizationRobotic IntelligenceOrdinary Differential Equation

🎯 What it does: Proposed a simplified model based on Moreau's stepping forward method, and validated its effectiveness in vertical swinging snake robot motion planning through simulation and experiments.

Optimization Based Human-Guided Variable-Stiffness Visual Impedance Control for Contact-Rich Tasks

Jiao Jiang, Hui Zhang

OptimizationRobotic IntelligenceImage

🎯 What it does: Proposed a human-guided visual impedance control framework for contact-rich tasks (such as polishing and drilling), and developed a variable stiffness visual impedance control strategy that uses online quadratic programming optimization to adjust impedance parameters, enabling the tool contact force to converge to the desired value while satisfying safety constraints.

Optimization-Based Path-Velocity Control for Time-Optimal Path Tracking under Uncertainties

Zheng Jia, Björn Olofsson

Autonomous DrivingOptimization

🎯 What it does: Proposed a real-time prediction scaling algorithm to achieve time-optimal path tracking under model uncertainty.

Optimized Optical Fiber Sensors for Forearm Muscle Deformation Monitoring and Hand Motion Recognition

Heifu Liu, Wei Meng

RecognitionPose EstimationTransformerBiomedical Data

🎯 What it does: Designed and optimized a flexible distributed fiber Bragg grating (FBG) sensor for monitoring forearm muscle deformation and hand motion recognition, and experimentally validated its performance.

Optimizing for Ride Comfort: A Model Predictive Control Framework with Frequency-Domain Analysis of the Acceleration Sequence

Chun-Chien Hsiao, Alireza Talebpour

Autonomous DrivingOptimizationSequential

🎯 What it does: Proposes an MPC framework centered on ride comfort, optimizing tangential and lateral acceleration patterns in 2D maneuvers, including longitudinal acceleration and steering rate.

Optoelectronic Navigation-Based Microtruck: For Efficient Cargo Loading, Transport, and Unloading

Ao Wang, Lin Feng

OptimizationPhysics Related

🎯 What it does: Propose a photonic navigation microcar based on Ag-SiO₂ microspheres, achieving efficient adsorption, rapid transportation, and targeted unloading of negatively charged dielectric particles by adjusting electric field frequency and optical parameters.

ORA-NET: Enhancing Image Feature Matching through Oriented Overlapping Region Alignment

Te Cui, Yufeng Yue

RetrievalImage

🎯 What it does: Proposes ORA-NET to enhance image feature matching performance through directional overlapping region alignment

ORBiT: Optimizing Robot-Assisted Bite Transfer Leveraging a Real2Sim2Real Framework

Sherwin Stephen Chan, Wei Tech Ang

OptimizationRobotic Intelligence

🎯 What it does: Developed the ORBiT framework, which optimizes bite transfer during robot-assisted feeding using motion capture-driven high-fidelity soft-body simulation, identifies transfer parameters that minimize contact force through systematic parameter search, and finally validates the approach on a real robot system.

ORCA: An Open-Source, Reliable, Cost-Effective, Anthropomorphic Robotic Hand for Uninterrupted Dexterous Task Learning

Clemens C. Christoph, Robert K. Katzschmann

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed and publicly released a 17-degree-of-freedom (DOF) tendon-driven humanoid robot hand ORCA, which can be assembled within 8 hours at a low cost (<2000 CHF), and demonstrated its applications in teleoperation, imitation learning, and zero-shot sim-to-real reinforcement learning tasks.

OrchardDepth++: Binned KL-Flood Regularization for Monocular Depth Estimation of Orchard Scene

Zhichao Zheng, Bruce A. MacDonald

Depth EstimationSupervised Fine-TuningImageAgriculture Related

🎯 What it does: Proposed and implemented a monocular depth estimation method based on binned KL-Flood regularization, specifically designed for orchard scenarios.

Origami-Inspired Pneumatic Continuum Manipulator: Stiffness Modeling and Validation

Zhuowen Li, Fan Xu

Robotic Intelligence

🎯 What it does: Developed a stiffness model for the schematic pneumatic continuum manipulator (OPM) and experimentally validated its predictive accuracy;

Origami-Inspired Soft Gripper with Tunable Constant Force Output

Zhenwei Ni, Cecilia Laschi

Robotic Intelligence

🎯 What it does: Proposed a flexible gripper based on origami principles that can achieve adjustable constant grip force within a wide deformation range;

Oscillation Suppression of Acoustic Trapping: A Disturbance Observer-based Approach

Yuyu Jia, Song Liu

OptimizationImagePhysics Related

🎯 What it does: Propose a control method based on visual feedback, which uses a binocular microscope vision system to locate particles and dynamically adjusts the sound field distribution through a disturbance observer, thereby suppressing the oscillation of the traveling-wave acoustic comb micro-sector laser along the z-axis.

OSMa-Bench: Evaluating Open Semantic Mapping Under Varying Lighting Conditions

M. Popov, S. Kolyubin

SegmentationVision Language ModelSimultaneous Localization and MappingVideoPoint CloudBenchmark

🎯 What it does: Evaluated the performance of open-source semantic mapping (OSM) under different indoor lighting conditions; proposed a dynamic configurable, automated LLM/LVLM-driven evaluation pipeline called OSMa-Bench; and introduced a new simulated RGB-D sequence dataset along with corresponding 3D reconstructed ground truth data for rigorous performance analysis.

Osmosis-Driven Large-Scale Actuation for Shape-Shifting Mechanisms

Elio J. Challita, Robert J. Wood

Robotic IntelligencePhysics Related

🎯 What it does: Achieving large-scale, reversible deformation-driven actuation using superabsorbent polymer (SAP) particles, demonstrating expansion-contraction motion generating approximately 10 N force through water absorption swelling, and integrating it into a deformable wheel to enable adaptive locomotion on land and in water.

Out-of-Distribution Recovery with Object-Centric Keypoint Inverse Policy for Visuomotor Imitation Learning

George Jiayuan Gao, Nadia Figueroa

Robotic Intelligence

🎯 What it does: Proposed an object keypoint-based recovery framework OCR to address OOD scenarios in visual kinematic policy learning.

Overlap-Aware Feature Learning for Robust Unsupervised Domain Adaptation for 3D Semantic Segmentation

Junjie Chen, Kemi Ding

SegmentationDomain AdaptationAdversarial AttackContrastive LearningPoint Cloud

🎯 What it does: Proposes a three-component framework, including a robustness evaluation model, reversible attention alignment module, and quality-guided contrastive memory bank, to address the issues of feature overlap and structural erosion in 3D point cloud semantic segmentation under unsupervised domain adaptation, while enhancing model robustness against adversarial attacks.

OVSG-SLAM: Open-Vocabulary Semantic Gaussian Splatting SLAM

Zhehang Liu, Yuwei Wu

Autonomous DrivingComputational EfficiencyVision Language ModelGaussian SplattingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose an open-vocabulary semantic SLAM framework OVSG-SLAM that combines multimodal perception with 3D Gaussian diffusion

P2 Explore: Efficient Exploration in Unknown Cluttered Environment with Floor Plan Prediction

Kun Song, Mingyu Ding

Autonomous DrivingOptimizationRobotic IntelligenceSimultaneous Localization and MappingImage

🎯 What it does: Propose FPUNet for efficiently predicting layouts in noisy indoor environments, extracting room segmentation, and constructing their topological connectivity, while optimizing the access order of predicted rooms to provide high-level exploration guidance;

PACR: Point-Axis Constraint Reasoning for Enhanced Robotic Manipulation with Dexterity and Compliance

Haowen Xiong, Jianxing Liu

OptimizationRobotic IntelligenceImagePoint CloudChain-of-Thought

🎯 What it does: Proposed the PACR framework, which jointly optimizes robot trajectories and damping curves, and employs a dual-agent Vision-Language Model for constraint generation and validation as well as an error backtracking mechanism;

PainDiffusion: Learning to Express Pain

Quang Tien Dam, Joo-Ho Lee

GenerationDiffusion modelVideoBiomedical Data

🎯 What it does: Proposed PainDiffusion, a diffusion model that generates natural pain expressions;

PANDAS: Prediction and Detection of Accurate Slippage

Teng Yan, Yang Zhang

Anomaly DetectionMultimodalityPhysics Related

🎯 What it does: Propose the PANDAS framework, integrating physics-informed multimodal spatiotemporal networks with probabilistic temporal reasoning modules for sliding detection and prediction.

PanopticSplatting: End-to-End Panoptic Gaussian Splatting

Yuxuan Xie, Yue Wang

SegmentationGaussian SplattingPoint Cloud

🎯 What it does: Proposed an end-to-end open-vocabulary semantic panoptic Gaussian splatting system (PanopticSplatting), achieving 2D instance mask cross-frame association-free upsampling, and enhancing 3D segmentation consistency through label mixing and label warping.

Parallel Transmission Aware Co-Design: Enhancing Manipulator Performance Through Actuation-Space Optimization

Rohit Kumar, Frank Kirchner

OptimizationRobotic Intelligence

🎯 What it does: Propose a collaborative design method for parallel mechanisms, optimizing the gear ratio and performing trajectory optimization in the execution space.

Parameter Selections and Applications for Soft Bellows Actuators (SBAs) with Various Performance Metrics

Wenjing Zou, Chao Wang

Optimization

🎯 What it does: This paper systematically investigates the effects of six key parameters on the load capacity, displacement efficiency, and bending resistance of soft pneumatic actuators (SBAs), and based on the findings, designs task-specific SBAs for applications such as high-load pneumatic grippers, high-efficiency displacement stages, and insect-like crawling robots.

Parameterized Motion Planning for Aerial Manipulators in Contact with Unstructured Surfaces

Zhixing Zhang, Yaonan Wang

OptimizationRobotic Intelligence

🎯 What it does: Proposed a sampling-based motion planning method called PCS-FMT* for continuous contact motion planning of drone manipulators on complex unstructured surfaces.

ParkDiffusion: Heterogeneous Multi-Agent Multi-Modal Trajectory Prediction for Automated Parking using Diffusion Models

Jiarong Wei, Abhinav Valada

Autonomous DrivingGraph Neural NetworkDiffusion modelMultimodality

🎯 What it does: Proposed the ParkDiffusion model for automatic parking, achieving prediction of vehicle and pedestrian trajectories

PartGrasp: Generalizable Part-level Grasping via Semantic-Geometric Alignment

Haoyang Lu, Yufeng Yue

Pose EstimationRobotic Intelligence

🎯 What it does: Proposed the PartGrasp method, achieving general and precise part-level grasping through semantic-geometric hierarchical alignment.

Partial Feedback Linearization Control of a Cable-Suspended Multirotor Platform for Stabilization of an Attached Load

Hemjyoti Das, Christian Ott

Robotic IntelligencePhysics Related

🎯 What it does: Propose a control method based on partial feedback linearization (PFL) for the stabilization control of suspended aerial platforms and their additional loads.

PathCluster: Pedestrian Group-Adaptive Social Navigation in Dense Crowds

Nihal Gunukula, Aniket Bera

Robotic IntelligenceSequential

🎯 What it does: Propose the PathCluster method, which improves social navigation for mobile robots in extremely crowded environments through group generation based on similar trajectories.

Pathfinder for Low-altitude Aircraft with Binary Neural Network

Kaijie Yin, Hui Kong

GenerationAutonomous DrivingConvolutional Neural NetworkTransformerImageMultimodalityPoint Cloud

🎯 What it does: Proposes an OSM generation method utilizing low-altitude aircraft equipped with aerial sensors, with the core being a binary dual-stream road segmentation model based on LiDAR and camera data to achieve efficient path finding and generate complete OSM maps.

PAVLM: Advancing Point Cloud based Affordance Understanding Via Vision-Language Model

Shang-Ching Liu, Jianwei Zhang

RecognitionTransformerPrompt EngineeringVision Language ModelPoint CloudBenchmark

🎯 What it does: By combining a geometry-guided propagation module with the hidden embeddings of a Large Language Model (LLM), the Vision-Language Model is utilized to enhance manipulability understanding of point clouds.

PB-MOT: Pose-aware Association Boosted Online 3D Multi-Object Tracking

Bo Pang, Liang Li

Object TrackingPose EstimationAutonomous DrivingPoint CloudBenchmark

🎯 What it does: Proposes the PB-MOT online 3D multi-target tracking framework, integrating pose-aware association to enhance tracking performance.

PC-SRIF: Preconditioned Cholesky-based Square Root Information Filter for Vision-aided Inertial Navigation

Tong Ke, Ryan C. DuToit

Autonomous DrivingOptimizationSimultaneous Localization and Mapping

🎯 What it does: Proposed a preconditioned Cholesky-based square root information filter (PC-SRIF) for solving linear systems in visual-inertial navigation systems (VINS).

PC2P: Multi-Agent Path Finding via Personalized-Enhanced Communication and Crowd Perception

Guotao Li, Yuhui Sun

OptimizationReinforcement Learning

🎯 What it does: Propose PC2P, a multi-agent path planning method combining Q-learning, utilizing personalized enhanced communication, congestion awareness, and regional deadlock-breaking strategies

PCGE: Boosting 3D Visual Grounding via Progressive Comprehension and Geometric-topology Perception Enhancement

Zeyu Wang, Jiamao Li

Pose Estimation

🎯 What it does: Proposes the PCGE framework, decomposing the 3D visual localization task into two steps: keypoint estimation and size regression.

PCMF2-Net: A Pyramid Cross-Modal Feature Fusion Network for Off-Road Freespace Detection

Ming Gao, Hui Kong

SegmentationAutonomous DrivingConvolutional Neural NetworkTransformerImageMultimodality

🎯 What it does: Proposes a pyramid cross-modal feature fusion network called PCMF2-Net for off-road airspace detection. The network takes dense depth maps, RGB images, and surface normal maps as input, employs a dual-branch CNN-Transformer encoder to extract local and global features, and uses a pyramid cross-modal feature fusion module for multi-scale, multi-modal feature fusion. It also incorporates an edge segmentation task and a two-step training strategy to enhance performance.

PD-VLA: Accelerating Vision-Language-Action Model Integrated with Action Chunking via Parallel Decoding

Wenxuan Song, Haoang Li

Computational EfficiencyRobotic IntelligenceVision-Language-Action Model

🎯 What it does: Proposed the PD-VLA framework, which accelerates the integration of Vision-Language-Action (VLA) models with action chunking through parallel decoding technology, thereby improving robotic manipulation efficiency.

PEACE: Prompt Engineering Automation for CLIPSeg Enhancement for Safe-Landing Zone Segmentation

Haechan Mark Bong, Giovanni Beltrame

SegmentationTransformerPrompt EngineeringVision Language ModelImage

🎯 What it does: Developed the PEACE system to automatically generate and refine CLIPSeg prompts for identifying safe landing zones in changing environments.

Peg-in-hole assembly method based on visual reinforcement learning and tactile pose estimation

Yong Tao, Hongxing Wei

Pose EstimationRobotic IntelligenceReinforcement LearningMultimodality

🎯 What it does: Proposes a pre-assembly method based on visual reinforcement learning combined with tactile pose estimation for real-time adjustment to improve the success rate of slot assembly.

Perception-aware Planning for Quadrotor Flight in Unknown and Feature-limited Environments

Chenxi Yu, Boyu Zhou

OptimizationRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposes a perception-driven planning method for quadrotors in unknown, feature-scarce environments, incorporating perspective conversion graphs and yaw trajectory generation that simultaneously balance exploration and localizability;

Performance consequences of information-based centralization arising from neural and mechanical coupling in a walking robot

Ellen Liu, P. Manoonpong

Robotic Intelligence

🎯 What it does: By adjusting joint stiffness and leg controller coupling, independently altering mechanical coupling and neural coupling to evaluate the degree of centralization in hexapod robots.

Perpetua: Multi-Hypothesis Persistence Modeling for Semi-Static Environments

Miguel A. Saavedra-Ruiz, Liam Paull

Object Tracking

🎯 What it does: Designed and implemented the Perpetua method to model the persistence and emergence probabilities of features in a semi-static environment, supporting multi-hypothesis tracking and predicting future feature states.

Persistent Preservation of a Spatio-temporal Environment Under Uncertainty

Amel Nestor Docena, Alberto Quattrini Li

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Studies how a single robot can continuously plan visit times and charge to maintain the spatiotemporal properties of regions of interest in the environment under state uncertainty;

Personalized Re-identification through Unsupervised Continual Learning and Parallel Training

Federico Rollo, Navvab Kashiri

RecognitionImage

🎯 What it does: Propose a method to enhance and continuously adapt re-identification for specific targets through unsupervised continual learning and intelligent image pool collection, personalizing neural networks.

Personalized Reinforcement Learning Control of Soft Robotic Exosuit for Assisting Human Normative Walking with Reduced Effort

Emiliano Quiñones Yumbla, Wenlong Zhang

Robotic IntelligenceSupervised Fine-TuningReinforcement LearningBiomedical Data

🎯 What it does: Propose an offline learning soft exoskeleton controller and online tuning to personalize assistance for human normal gait.

Personalized Robotic Achilles Tendon Utilizing a Semi-Passive Spring with Switching Stiffness*

Mingyu Seong, Jungsu Choi

Robotic Intelligence

🎯 What it does: Developed a lightweight semi-passive spring ankle assistive device called the robotic Achilles tendon (RAT), achieving elastic support during gait phases via double-acting cylinders and electromagnetic valves;

Pet-NODE Modeling: Embedding Priors and Time-Series Features into Neural ODE

Jia Chen, Shihua Li

Robotic IntelligenceTime SeriesOrdinary Differential Equation

🎯 What it does: Propose the Pet-NODE framework, which integrates physical priors and temporal features into a neural ordinary differential equation (NODE) model for high-fidelity modeling of mobile robot dynamic systems; and further embeds domain knowledge through a self-predictive target loss function.

PhyGrasp: Generalizing Robotic Grasping with Physics-informed Large Multimodal Models

Dingkun Guo, Wei Zhan

Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelTextMultimodalityPoint CloudPhysics Related

🎯 What it does: Proposed a physics-informed large-scale multimodal model called PhyGrasp, which integrates natural language and 3D point clouds through a bridging module to enhance robotic grasping performance; simultaneously constructed the PhyPartNet dataset containing 195K object instances and their linguistic descriptions, and conducted experimental evaluations on both simulated and real robots.

PhysGCN-DL: Physics-Informed Graph Convolutional Networks with Diversity-Aware Loss Optimization for Multimodal Pedestrian Trajectory Prediction

Zihan Jiang, Weidong Zhang

OptimizationGraph Neural NetworkMultimodalityPhysics Related

🎯 What it does: Proposed a physics-informed graph convolutional network (PhysGCN-DL), which models pedestrian dynamic interactions by representing physical interactions between pedestrians as edge weights in graph convolutions, thereby enhancing the diversity and accuracy of trajectory prediction.

Physical Human-Robot Collaboration-Assisted Acetabular Preparation for Total Hip Replacement Surgery

Ziqi Wang, Shoudong Huang

Robotic Intelligence

🎯 What it does: Using collaborative robots with variable impedance control for acetabular preparation in hip replacement surgery;

Physically-Feasible Reactive Synthesis for Terrain-Adaptive Locomotion via Trajectory Optimization and Symbolic Repair

Ziyi Zhou, Ye Zhao

OptimizationRobotic IntelligencePhysics Related

🎯 What it does: Propose an integrated planning framework for quadruped robots to achieve feasible ground gaits on dynamic, unpredictable terrains; the framework combines symbolic reactive synthesis with mixed-integer convex programming (MICP) for dynamic and physically feasible motion planning, and enhances the scalability of synthesis through symbolic repair and a high-level manager.