arXivSub Start free trial

IROS 2025 Papers — Page 10

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

Hybrid Learning-based Balance Function Assessment of Stroke Patients with a Single Ear-Worn IMU

Tianshu Zhao, Yao Guo

Convolutional Neural NetworkRecurrent Neural NetworkTime SeriesBiomedical Data

🎯 What it does: Propose a hybrid learning framework based on single-ear IMU for predicting Berg Balance Scale (BBS) scores in stroke patients during the 3-meter TUG test.

Hybrid Motion Control of a Fiber-based Soft Robotic Instrument for Minimally Invasive Surgery

Ziqi Yang, Burak Temelkuran

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Developed and verified a two-stage hybrid motion control framework for flexible surgical instruments based on thermoplastic fiber and tendon-driven mechanisms, integrating a learning-based inverse kinematics model with closed-loop feedback to achieve high-precision large-scale motion.

Hybrid Transformer-Mamba Model for 3D Semantic Segmentation

Xinyu Wang, Yingying Zhu

SegmentationTransformerPoint Cloud

🎯 What it does: Propose the HybridTM architecture, a hybrid of Transformer and Mamba for 3D semantic segmentation, and introduce the Inner Layer Hybrid Strategy to fuse the two at a finer-grained level to simultaneously capture long-range dependencies and fine-grained features.

Hyla-SLAM: Toward Maximally Scalable 3D LiDAR-Based SLAM Using Dynamic Memory Management and Behavior Trees

Steven Swanbeck, Mitchell W. Pryor

Autonomous DrivingRobotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes the Hyla-SLAM framework, integrating dynamic memory management and behavior trees to achieve efficient map creation and management for 3D LiDAR SLAM in environments of any scale.

Hyperbolic Transformers with LLMs for Multimodal Human Activity Recognition

F. Soleimani, Y. Amirat

RecognitionTransformerLarge Language ModelSupervised Fine-TuningMultimodalityTabular

🎯 What it does: Propose a strategy based on hyperbolic space optimization, integrating it into Transformer and GPT-2 models, and fine-tuning on both unimodal (UCI-HAR, Opportunity) and multimodal (UTD-MHAD, NTU RGB+D) datasets for human activity recognition.

HyperGraph ROS: An Open-Source Robot Operating System for Hybrid Parallel Computing based on Computational HyperGraph

Shufang Zhang, Shan An

Computational EfficiencyRobotic IntelligenceGraph

🎯 What it does: Developed the HyperGraph ROS system, unifying cross-process, cross-device, and cross-process computing into a computational hypergraph to enhance message passing and parallel execution efficiency.

ICCO: Learning an Instruction-conditioned Coordinator for Language-guided Task-aligned Multi-robot Control

Yoshiki Yano, Takamitsu Matsubara

Robotic IntelligenceReinforcement LearningText

🎯 What it does: Proposed the ICCO framework, achieving language-guided multi-robot task alignment control through multi-agent reinforcement learning;

IDAGC: Adaptive Generalized Human-Robot Collaboration via Human Intent Estimation and Multimodal Policy Learning

Haotian Liu, Zhengtao Zhang

Robotic IntelligenceTransformerAuto EncoderMultimodality

🎯 What it does: Proposed an intent-driven adaptive collaboration framework IDAGC based on multimodal data, achieving automatic recognition and switching of human-robot collaboration modes, and learning multi-task strategies.

iGaussian: Real-Time Camera Pose Estimation via Feed-Forward 3D Gaussian Splatting Inversion

Hao Wang, Haibin Yan

Pose EstimationGaussian SplattingImage

🎯 What it does: Proposes iGaussian, a two-stage feedforward framework that realizes real-time camera pose estimation through direct 3D Gaussian inversion.

IGDrivSim: A Benchmark for the Imitation Gap in Autonomous Driving

Clémence Grislain, Shimon Whiteson

Autonomous DrivingReinforcement LearningBenchmark

🎯 What it does: Developed and introduced the IGDrivSim benchmark to study the imitation learning gap caused by differences between human perception and vehicle sensors in simulated environments, and experimentally verified the impact of this gap on learning safe and effective driving behaviors; further demonstrated that combining imitation learning with reinforcement learning and using penalty rewards for forbidden behaviors can effectively mitigate such failures.

IHGSL: Interpretable Heuristic Graph Structure Learning for Multi-Robot Autonomous Collaborative Systems

Yue Han, Zhixiao Sun

Explainability and InterpretabilityRobotic IntelligenceGraph Neural NetworkGraph

🎯 What it does: Proposes an interpretable heuristic graph structure learning (IHGSL) method to better understand complex collaborative relationships in multi-robot systems;

Image-Based Relocalization and Alignment for Long-Term Monitoring of Dynamic Underwater Environments

Beverley Gorry, Alejandro Fontan

RetrievalSimultaneous Localization and MappingImageBenchmark

🎯 What it does: Developed an image-based relocalization and alignment pipeline for long-term monitoring of dynamic underwater environments

Image-Goal Navigation Using Refined Feature Guidance and Scene Graph Enhancement

Zhicheng Feng, Huimin Lu

Autonomous DrivingKnowledge DistillationImage

🎯 What it does: Proposes an image goal navigation method called RFSG, which leverages fine-grained associations between goals, observations, and environments in limited image data while maintaining a lightweight and concise architecture.

Imagine-2-Drive: Leveraging High-Fidelity World Models via Multi-Modal Diffusion Policies

Anant Garg, K. Krishna

Autonomous DrivingReinforcement LearningDiffusion modelWorld ModelMultimodality

🎯 What it does: Proposed a WMRL framework that combines high-fidelity world models with multi-modal diffusion strategies, enabling robust driving policy learning with minimal online interaction

Imitation-Guided Bimanual Planning for Stable Manipulation under Changing External Forces

Kuanqi Cai, Arash Ajoudani

Robotic Intelligence

🎯 What it does: Proposes an imitation-based bimanual planning framework that integrates efficient grasping transition strategies and motion performance optimization; introduces a strategy of sampling stable intersection points in the grasping manifold to achieve seamless transition from single-handed grasping to bimanual grasping; constructs a hierarchical two-stage motion architecture that combines the global path generator from imitation learning with a local planner driven by quadratic programming (QP), ensuring real-time motion feasibility, obstacle avoidance, and improved controllability.

IMM-MOT: A Novel 3D Multi-object Tracking Framework with Interacting Multiple Model Filter

Xiao-He Liu, Yuhan Wang

Object TrackingAutonomous DrivingScore-based ModelPoint CloudBenchmark

🎯 What it does: Proposed a 3D multi-object tracking framework called IMM-MOT based on the Interacting Multiple Model (IMM) filter, introducing the Damping Window mechanism and the Distance-Based Score Enhancement module to improve trajectory lifecycle management and detection score processing.

IMMNN: Robust Wireless Electromagnetic-Inertial Fusion Tracking via Learning An Adaptive IMM

Sichao Lin, Tim C. Lueth

Object TrackingRecurrent Neural NetworkGraph Neural NetworkMultimodality

🎯 What it does: Propose a learning-based Interacting Multiple Model (IMMNN) wireless electromagnetic-inertial fusion tracking system, combining a multi-transmitter array with a WEMT-IMU fusion tracker;

Impact of Heterogeneous UWB Sensor Noise on the Optimality and Sensitivity of Mobile Positioning Systems

Mathilde Theunissen, E. Malis

OptimizationRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed a theoretical framework for designing multi-robot formations equipped with UWB sensors and localizing target robots in the presence of noisy distance measurements.

Impact of Static Friction on Sim2Real in Robotic Reinforcement Learning

Xiaoyi Hu, Jiangwei Zhong

Domain AdaptationRobotic IntelligenceReinforcement Learning

🎯 What it does: Study the impact of static friction on Sim2Real, propose a static friction-aware domain randomization method, and verify its performance in Sim2Sim and Sim2Real experiments using a control theory joint model for parameter identification, which outperforms traditional domain randomization and Actuator Net.

Impact of Temporal Delay on Radar-Inertial Odometry

Vlaho-Josip Štironja, Ivan Petrović

Autonomous DrivingOptimizationSimultaneous Localization and MappingMultimodality

🎯 What it does: Proposes a radar-inertial odometry (RIO) system that fuses automotive radar with an inertial measurement unit (IMU) and integrates online delay calibration in factor graph optimization.

ImpedanceGPT: VLM-driven Impedance Control of Swarm of Mini-drones for Intelligent Navigation in Dynamic Environment

Faryal Batool, D. Tsetserukou

Robotic IntelligenceTransformerVision Language ModelRetrieval-Augmented Generation

🎯 What it does: Proposes ImpedanceGPT, a system based on vision-language models (VLM) and retrieval-augmented generation (RAG) frameworks, for achieving impedance control and adaptive navigation of small drone swarms in dynamic environments.

Implicit Disparity-Blur Alignment for Fast and Precise Autofocus in Robotic Microsurgical Imaging

Pan Fu, Gui-Bin Bian

Depth EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Propose an implicit disparity-blur alignment method to achieve rapid and precise auto-focusing in robotic microsurgery, with prototype verification.

Improbability Roller-2: A Hybrid Mobile Robot with Variable Diameter Transformable Wheels

Gourav Moger, H. A. Varol

Robotic Intelligence

🎯 What it does: Developed a hybrid mobile robot named Improbability Roller 2 with variable diameter wheels, simplifying the design and enhancing mobility and adaptability on irregular terrain.

Improved 2D Hand Trajectory Prediction with Multi-View Consistency

Junyi Ma, Hesheng Wang

GenerationPose EstimationDiffusion modelSequential

🎯 What it does: Propose and implement a learning scheme based on EER to leverage multi-view consistency and enhance the accuracy of a diffusion-based 2D hand trajectory prediction method.

Improved 3D Point-Line Mapping Regression for Camera Relocalization

B. Bui, Joo-Ho Lee

Pose EstimationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a novel architecture that independently learns features for 3D points and lines, then combines them for camera relocalization.

Improved Calibration for Panoramic Annular Lens Systems with Angular Modulation

Ding Wang, Lingbao Kong

Image

🎯 What it does: To address the calibration challenges of panoramic mirror systems, a new projection model incorporating angular modulation is proposed and validated on both synthetic and real datasets.

Improved Free Motion Performance for TDPA-Passivated Position-Force Measured Teleoperation Architectures

Camilla Celli, A. Frisoli

Robotic Intelligence

🎯 What it does: Proposes a new form that does not require passivization via TDPA during free motion, provides a formal proof, and demonstrates its effectiveness using a one-degree-of-freedom case.

Improving 6D Object Pose Estimation of Metallic Household and Industry Objects

Thomas Pöllabauer, A. Kuijper

Pose EstimationImageBenchmark

🎯 What it does: Create a new BOP-compatible metal object dataset and improve the GDRNPP algorithm on this dataset by adding extra keypoint prediction and material estimation heads to enhance 6D pose estimation accuracy.

Improving Drone Racing Performance Through Iterative Learning MPC

Haochen Zhao, Angela P. Schoellig

Autonomous DrivingOptimization

🎯 What it does: Improved iterative learning model predictive control (LMPC) to enhance drone racing performance.

Improving Generalization of Language-Conditioned Robot Manipulation

Chenglin Cui, Andrea Cavallaro

Robotic IntelligenceVision Language ModelVision-Language-Action ModelImageText

🎯 What it does: Proposes a two-stage framework to learn object arrangement tasks with a few demonstrations, achieving target recognition and localization based on natural language instructions through an instance-level semantic fusion module.

Improving Visual Place Recognition with Sequence-Matching Receptiveness Prediction

Somayeh Hussaini, Michael Milford

RetrievalImage

🎯 What it does: Propose a supervised learning method to predict the acceptability (SMR) of each frame's sequence matching, enabling the system to decide when to trust the sequence matching results.

IMU-Based Motion Mode Recognition in Soft Underwater Exosuit

Mengbo Luan, Xinyu Wu

RecognitionRobotic IntelligenceConvolutional Neural NetworkRecurrent Neural NetworkTime Series

🎯 What it does: Proposed an LSTM-CNN-based soft underwater exoskeleton motion pattern recognizer capable of performing motion pattern classification and state transition label identification.

IMVPR: Implicit BEV-Enhanced Multi-View Aggregation for Visual Place Recognition

Xu Cao, Hao Fang

RetrievalAutonomous DrivingImagePoint Cloud

🎯 What it does: Proposed the IMVPR network, achieving visual localization through implicit BEV-enhanced multi-view feature fusion and descriptor aggregation.

In-Situ Classification of Soil Types Exploiting Electrical Impedance Tomography with a Robotic Actuating Probe

Xiaoxian Xu, Fumiya Iida

ClassificationRobotic IntelligenceAgriculture Related

🎯 What it does: Proposed a method that utilizes electrical resistance tomography (EIT) combined with a movable probe to achieve on-site soil type classification.

In-situ Value-aligned Human-Robot Interactions with Physical Constraints

Hongtao Li, Zilong Zheng

Robotic IntelligenceLarge Language ModelBenchmark

🎯 What it does: Propose a framework that combines human preferences with physical constraints, and develop a benchmark for daily household activities, generating task plans through In-Context Learning from Human Feedback (ICLHF) derived from direct instructions and adjustments in daily life.

Incremental Language Understanding for Online Motion Planning of Robot Manipulators

Mitchell Abrams (Technische Universität Wien), M. Scheutz (Technische Universität Wien)

Robotic Intelligence

🎯 What it does: Proposed a cognitive architecture that integrates online motion planning algorithms with incremental language parsers for incremental language understanding and real-time motion planning in robotic manipulators.

Incremental Learning for Robot Shared Autonomy

Yiran Tao, Zackory Erickson

Robotic IntelligenceSupervised Fine-Tuning

🎯 What it does: Proposes the ILSA (Incrementally Learned Shared Autonomy) framework, which utilizes user interaction to continuously improve the assistance strategies of robots in shared autonomy.

Indoor FireRescue Radar: 4D Indoor Millimeter Wave Dataset and Analysis for Hazardous Environment Perception

Kangkang Duan, Zhengbo Zou

Object DetectionMultimodalityPoint CloudBenchmark

🎯 What it does: Proposed and released the Indoor FireRescue Radar (IFR) dataset, and conducted experiments on voxel and pillar-based detectors for 4D radar point clouds to verify the robustness of radar in fire environments.

Inducing Desired Equilibria in Constrained Noncooperative Games via Nudging

Ao Wang, Xiuxian Li

OptimizationRobotic Intelligence

🎯 What it does: Designed and analyzed a central regulator's nudge strategy, aiming to guide players in restricted non-cooperative games toward an expected equilibrium through trust variables and distributed updates, and applied it to robot formation control.

Industrial-Grade Sensor Simulation via Gaussian Splatting: A Modular Framework for Scalable Editing and Full-Stack Validation

Xianming Zeng, Qiang Li

Autonomous DrivingDiffusion modelGaussian SplattingImagePoint Cloud

🎯 What it does: Proposed an industrial-grade sensor simulation framework based on Gaussian Splatting, and reconstructed three key components to achieve explicit scene representation and real-time rendering.

Industry 6.0: New Generation of Industry driven by Generative AI and Swarm of Heterogeneous Robots

Artem Lykov, D. Tsetserukou

Robotic IntelligenceTransformerLarge Language ModelText

🎯 What it does: Proposed and implemented the Industry 6.0 concept, which is a fully automated production system capable of autonomously completing the entire product design and manufacturing process based on natural language descriptions provided by users.

Information Entropy-assisted Hierarchical Framework for Unknown Environments Exploration

Changjun Gu, Xinbo Gao

OptimizationRobotic Intelligence

🎯 What it does: Proposed an information entropy-assisted hierarchical planning framework (IEHP) for efficient autonomous exploration, designed an efficient sub-region arrangement method considering total travel distance, path similarity, and information entropy, and introduced a globally consistent frontier selection method to reduce redundant local paths and enhance consistency between local and global planning.

Inspection Planning Primitives with Implicit Models

Jingyang You, Lashika Medagoda

Computational EfficiencyRobotic IntelligenceMesh

🎯 What it does: Proposed a set of primitive computations called IPIM, enabling sampling-based inspection planners to fully utilize neural SDF representations for planning.

Inspiring External Human-Machine Interface Designs for Autonomous Personal Mobility Vehicle: Causal Discovering the Influence of Passengers’ Personality Traits on User Experience

Hailong Liu, Takahiro Wada

Autonomous Driving

🎯 What it does: Studied the impact of passenger personality traits on user experience when using different external human-machine interfaces (eHMI), and revealed their relationship through field experiments and causal discovery analysis;

Instantaneous Contact Localization on A Magnetically Transduced Tapered Whisker

Yixuan Dang, A. Knoll

OptimizationRobotic Intelligence

🎯 What it does: Developed a magnetic sensing conical hair-based tactile sensor integrated with axial force measurement, capable of distinguishing tangential contact and localizing contact points.

Instantaneous Walkability Determination Method for Almost Linear Passive Dynamic Walker with Nontrivial Limit Cycle Stability

Fumihiko Asano, Taiki Sedoguchi

OptimizationRobotic IntelligencePhysics RelatedOrdinary Differential Equation

🎯 What it does: Propose an instantaneous walkability determination method for near-linear passive dynamic walkers, whose gait is period 1 and asymptotically stable, but limit cycle stability is non-trivial.

Integrating Ergonomics and Manipulability for Upper Limb Postural Optimization in Bimanual Human-Robot Collaboration

Chenzui Li, Fei Chen

Optimization

🎯 What it does: Propose an upper limb posture optimization method that enhances ergonomics safety and force manipulability in dual-arm collaborative object transportation tasks by minimizing a cost function, generating robot end-effector reference postures through a transformation module, and achieving posture recalibration via a damping controller (MPIC) combined with a dual-arm model.

Integrating Offline Pre-Training with Online Fine-Tuning: A Reinforcement Learning Approach for Robot Social Navigation

Run Su, Ying Fu

Robotic IntelligenceTransformerReinforcement Learning

🎯 What it does: Proposed an RL algorithm that combines offline pre-training with online fine-tuning, integrating RTG prediction into a causal Transformer to achieve robot social navigation;

Integrating Opinion Dynamics into Safety Control for Decentralized Airplane Encounter Resolution

Shuhao Qi, S. Haesaert

🎯 What it does: Integrate bio-inspired nonlinear opinion dynamics into a decentralized aircraft collision avoidance safety control framework to ensure safe and unobstructed solutions.

Integrating Trajectory Optimization and Reinforcement Learning for Quadrupedal Jumping with Terrain-Adaptive Landing

Renjie Wang, Donglin Wang

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a safety landing framework that combines trajectory optimization with reinforcement learning, enabling quadruped robots to achieve adaptive landing on rough terrain

Intelligent LiDAR Navigation: Leveraging External Information and Semantic Maps with LLM as Copilot

Fujing Xie, Sören Schwertfeger

Autonomous DrivingRobotic IntelligenceLarge Language ModelTextPoint Cloud

🎯 What it does: Propose combining osmAG (a semantic topological hierarchical map based on OpenStreetMap text) with a large language model (LLM) as a co-pilot for LiDAR navigation to achieve richer external information fusion;

Intensity-Augmented LiDAR-Visual-Inertial Odometry and Meshing

YunFeng Hua, WeiWei Xu

Autonomous DrivingOptimizationSimultaneous Localization and MappingImagePoint CloudMesh

🎯 What it does: Developed a tightly coupled LiDAR-Visual-Inertial Odometry system that integrates LIO and VIO, using point-to-mesh tracking and LiDAR intensity information to optimize pose, and achieving real-time TSDF mapping and Marching Cubes grid extraction on the GPU.

Interactive Expressive Motion Generation Using Dynamic Movement Primitives

Till Hielscher, K. Arras

GenerationRobotic IntelligenceOrdinary Differential Equation

🎯 What it does: Proposes a framework that utilizes Dynamic Movement Primitives (DMP) to implement 12 animation principles for automatically generating robot motions with socially interactive emotional expressions.

Interactive Fine-grained Few-shot Detection of Tools*

Philipp Keller, R. Dillmann

ClassificationObject DetectionRobotic IntelligenceTransformerImage

🎯 What it does: Proposed a DE-fine-ViT model that does not require fine-tuning for fine-grained few-shot tool detection in mobile robot scenarios, where users can construct class and component prototypes during the interactive preparation phase, and during inference, a re-evaluation module utilizes multi-grained prototypes to achieve fine-grained classification.

Interactive Navigation for Legged Manipulators with Learned Arm-Pushing Controller

Zhihai Bi, Jun Ma

Robotic IntelligenceReinforcement Learning

🎯 What it does: An interactive navigation framework for legged robots' arms is proposed, utilizing an active arm-pushing mechanism to reposition movable obstacles in confined spaces; a dual-stage reward-based arm-pushing controller is designed using reinforcement learning.

Interactive Object Detection by Mitigating Uncertainty of Robot Task Plans using Large Language Model

Kanata Suzuki, Tetsuya Ogata

Object DetectionRobotic IntelligenceLarge Language Model

🎯 What it does: Proposed an interactive task planning method based on large language models, which considers task goals during object recognition and clarifies the robot's task purpose through question-answering, thereby updating detection queries to align with task goals.

Interdigitated Electrodes for Selective Stimulation of Skeletal Muscle Actuators in Biosyncretic Robots

Lianchao Yang, Lianqing Liu

Robotic IntelligencePositron Emission Tomography

🎯 What it does: Selective stimulation of 3D skeletal muscle tissue using thin-film interdigitated electrodes (IDEs)

Interpretable Interaction Modeling for Trajectory Prediction via Agent Selection and Physical Coefficient

Shiji Huang, Deyuan Liang

Autonomous DrivingExplainability and InterpretabilityComputational EfficiencyTransformerSequential

🎯 What it does: Proposed the ASPILin method, which achieves interpretable interaction modeling by manually selecting interacting agents and replacing the attention scores in Transformer with physical correlation coefficients, thereby improving the accuracy of trajectory prediction.

Interpreting Behaviors and Geometric Constraints as Knowledge Graphs for Robot Manipulation Control

Chen Jiang, Martin Jägersand

Explainability and InterpretabilityRobotic IntelligenceGraph Neural Network

🎯 What it does: Explores the use of knowledge graphs to explain and control robot operational behaviors, and unifies behavior trees with geometric constraints

Inverse Attention-Weighted Model with Heterogeneous Spatio-Temporal Interaction Graph for Autonomous Navigation Systems

Yi-Lin Li, Hsiao-Ping Tsai

Autonomous DrivingRobotic IntelligenceGraph Neural NetworkTransformerGraph

🎯 What it does: Propose a Res-Mlp based attention mechanism and inverse attention weighting module to enhance robot navigation and collision avoidance in environments coexisting with crowds and obstacles.

Inverse Kinematics for Robot Arm Using Minimum Mean Square Error

C. Shin, Hoseong Kwak

OptimizationRobotic Intelligence

🎯 What it does: Studied the use of minimum mean square error and variance-based control methods to solve the inverse kinematics problem of robotic arms.

Inverse-Free and Data-Driven Motion Tracking Control for Redundant Robot with Fuzzy Recurrent Neural Network

Min Yang, Hui Zhang

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Proposed a data-driven fuzzy discrete recurrent neural network (D2-FDRNN) model to address the precise motion tracking control problem for redundant robots under unknown structural knowledge and noise interference.

Investigating the Fitness of Finger Grippers for Dynamic Tactile Manipulation Under Static Object Conditions

M. C. Yildirim, Sami Haddadin

Robotic Intelligence

🎯 What it does: Proposed an integrated framework that unifies separated and coupled gripper metrics into a single perspective, introduced 16 evaluation indicators for gripper force control, force response, and efficiency, designed three new experimental setups to quantify these indicators, and evaluated three commercial finger grippers.

Investigating the Impact of Humor on Learning in Robot-Assisted Education

Xiaoxuan Hei, Adriana Tapus

Robotic Intelligence

🎯 What it does: Investigate the impact of humor in robot-assisted education by comparing the effects of three humor strategies (no humor, preset timing humor, adaptive timing humor) on students' learning outcomes and overall learning experience.

IoU-Aware Clustering for Anchor Configuration Determination in Efficient Defect Detection

Yuhao Zhao, Song Liu

Object DetectionImage

🎯 What it does: Proposes the IoU-Aware Clustering algorithm for automatically learning anchor box configurations in surface defect detection.

Is the House Ready For Sleeptime? Generating and Evaluating Situational Queries for Embodied Question Answering

Vishnu Sashank Dorbala, Dinesh Manocha

Data SynthesisRobotic IntelligenceLarge Language ModelPrompt EngineeringText

🎯 What it does: Proposed and implemented the embodied question answering (S-EQA) task for contextual queries in home environments, generating and evaluating contextual queries and object states using the Prompt-Generate-Evaluate (PGE) method.

Iterative Camera-LiDAR Extrinsic Optimization via Surrogate Diffusion

Ni Ou, Junzheng Wang

OptimizationDiffusion modelImagePoint Cloud

🎯 What it does: Proposes an iterative framework based on proxy diffusion to progressively refine initial extrinsic parameters in camera and LiDAR extrinsic parameter calibration through repeated denoising processes, thereby enhancing the performance of any existing calibration method.

Iterative Learning Motion Control of Continuum Robots Based on Neural Ordinary Differential Equations

Zhenhan Liang, Ning Tan

Robotic IntelligenceOrdinary Differential Equation

🎯 What it does: Proposed a data-driven iterative learning control system based on Neural Ordinary Differential Equations (NODE) for continuum robots.

iWalker: Imperative Visual Planning for Walking Humanoid Robot

Xiaodi Lin, Chen Wang

OptimizationRobotic Intelligence

🎯 What it does: Proposed an end-to-end visual perception-planning-execution walking system called iWalker, achieving unified control of visual obstacle avoidance, gait planning, and whole-body balance.

J-ORA: A Framework and Multimodal Dataset for Japanese Object Identification, Reference, Action Prediction in Robot Perception

Jesse Atuhurra, Koichiro Yoshino

RecognitionRobotic IntelligenceVision Language ModelVision-Language-Action ModelMultimodality

🎯 What it does: Proposed the J-ORA multimodal dataset and evaluated visual language models (VLMs) on object recognition, coreference resolution, and next-action prediction tasks.

Jacobian Exploratory Dual-Phase Reinforcement Learning for Dynamic Endoluminal Navigation of Deformable Continuum Robots

Yu Tian, Hongliang Ren

Robotic IntelligenceReinforcement LearningBiomedical Data

🎯 What it does: Proposed the Jacobian Exploratory Dual-Phase Reinforcement Learning (JEDP-RL) framework for deformable continuum robots in dynamic cavity navigation

JAM: Keypoint-Guided Joint Prediction after Classification-Aware Marginal Proposal for Multi-Agent Interaction

Fangze Lin, Hong Zhang

Autonomous DrivingPoint CloudBenchmark

🎯 What it does: Proposed a two-stage multi-agent interaction prediction framework named JAM, which first performs classification-aware edge prediction and then completes joint prediction under the guidance of key points to address the issue of low-quality generation of low-probability patterns.

JENGA: Object selection and pose estimation for robotic grasping from a stack

Sai Srinivas Jeevanandam, J. Rambach

Pose EstimationRobotic IntelligenceImage

🎯 What it does: Propose a camera-IMU based method for selecting graspable objects in stacked structures and estimating their 6DoF poses, while introducing corresponding datasets and evaluation metrics.

JiAo: A Versatile Snake Robot with Elliptical Wheels for Multimodal Locomotion

Zizhu Zhao, Shiyong Meng

Robotic Intelligence

🎯 What it does: Designed and built a multi-modal snake robot named JiAo with elliptical wheels, developed corresponding wheeled and body-based locomotion control systems, and conducted experimental validation in various complex environments.

Joint Optimization of Multi-Agent Task Allocation and Path Planning for Continuous Pickup and Delivery Tasks

Hongkai Fan, Qin Tan

Optimization

🎯 What it does: Proposes a task allocation method based on cost matrices and MILP, and builds the CBS-TAPF framework to achieve joint optimization of task allocation and path planning, while extending it to continuous task scenarios, enhancing robustness and adaptability through dynamic updates of the task allocation matrix queue.

Joint Pedestrian and Vehicle Traffic Optimization in Urban Environments using Reinforcement Learning

Bibek Poudel, K. Heaslip

OptimizationReinforcement LearningVideo

🎯 What it does: Propose a deep reinforcement learning framework for adaptive control of eight traffic signals, jointly optimizing the efficiency of pedestrians and vehicles.

JPDS-NN: Reinforcement Learning-Based Dynamic Task Allocation for Agricultural Vehicle Routing Optimization

Yixuan Fan, Tao Zhang

OptimizationGraph Neural NetworkTransformerReinforcement LearningAgriculture Related

🎯 What it does: Proposed a Joint Probability Distribution Sampling Neural Network (JPDS-NN) for solving the Entry-Dependent Vehicle Routing Problem (EDVRP) in agricultural scenarios, achieving end-to-end rapid planning through reinforcement learning.

Jumping Mechanism Assists Takeoff for Large-Sized Flapping-Wing Robots

Zeyu Zhang, Pengfei Zhang

Robotic Intelligence

🎯 What it does: Designed a jumping mechanism based on an arch-shaped carbon fiber spring to assist large winged robots in takeoff, and proposed a jumping-flapping wing coupling control method, verifying its effectiveness.

Kalib: Easy Hand-Eye Calibration with Reference Point Tracking

Tutian Tang, Cewu Lu

Robotic IntelligenceVision-Language-Action Model

🎯 What it does: Propose the Kalib method to achieve automatic and easy-to-setup hand-eye calibration using the robot's kinematic chain and predefined reference points.

KARL: Kalman-Filter Assisted Reinforcement Learner for Dynamic Object Tracking and Grasping

Kowndinya Boyalakuntla, Jingjin Yu

Object TrackingRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposes the KARL system for dynamic object tracking and grasping in eye-in-hand (EoH) systems.

KD-RIEKF: Kinodynamic Right-Invariant EKF for Legged Robot State Estimation

Qi Yang, Bin Liang

Robotic IntelligencePhysics Related

🎯 What it does: Propose the KD-RIEKF framework, integrating dynamic constraints into the right-invariant extended Kalman filter (RIEKF) for state estimation in legged robots

KDMOS:Knowledge Distillation for Motion Segmentation

Chunyu Cao, Xiaoyu Tang

SegmentationAutonomous DrivingKnowledge DistillationConvolutional Neural NetworkPoint Cloud

🎯 What it does: Proposes a logits-based knowledge distillation framework, using a bird's eye view (BEV) projection model as the student, a non-projection model as the teacher, and separating moving and non-moving categories. Custom distillation strategies, dynamic upsampling, and network structure optimization are employed to improve accuracy while maintaining real-time inference speed.

Keypoint-Aware RAG for Robotic Manipulation: In-Context Constraint Learning via Large-Scale Retrieval

Jiuzhou Lin, Houde Liu

RetrievalRobotic IntelligenceVision Language ModelRetrieval-Augmented Generation

🎯 What it does: Propose Keypoint-Aware Retrieval Augmented Generation (KARAG), achieving retrieval-augmented generation between vision-language models and robot datasets through keypoint constraints;

KGN-Pro: Keypoint-Based Grasp Prediction through Probabilistic 2D-3D Correspondence Learning

Bingran Chen, Guangyao Zhai

Pose EstimationRobotic IntelligenceImage

🎯 What it does: Proposed a grasp prediction network named KGN-Pro, which achieves keypoint-based 6-DoF grasp prediction through probabilistic 2D-3D correspondence learning.

KineDepth: Utilizing Robot Kinematics for Online Metric Depth Estimation

Simranjeet Singh, Michael C. Yip

Depth EstimationRobotic IntelligenceRecurrent Neural NetworkImage

🎯 What it does: Using a single camera combined with robot kinematics, real-time conversion of monocular relative depth estimation into metric depth;

Kinematic Model and Trajectory Tracking Algorithm for High-Speed Spherical Robots *

Bixuan Zhang, Guang Li

OptimizationRobotic Intelligence

🎯 What it does: Proposed a novel theory for turning of spherical robots, and based on this theory, constructed and optimized a new kinematic model, followed by designing a trajectory tracking algorithm that maintains reliable performance under high-speed conditions.

KISS-SLAM: A Simple, Robust, and Accurate 3D LiDAR SLAM System With Enhanced Generalization Capabilities

Tiziano Guadagnino, C. Stachniss

Autonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Implemented a LiDAR-only SLAM system that is concise, robust, and accurate.

KiteRunner: Language-Driven Cooperative Local-Global Navigation Policy with UAV Mapping in Outdoor Environments

Shibo Huang, Xian Wei

OptimizationRobotic IntelligenceTransformerLarge Language ModelVision Language ModelDiffusion modelImageTextMultimodality

🎯 What it does: Propose the KiteRunner framework, combining global planning using drone orthographic images and local path generation driven by diffusion models, to achieve language-driven collaborative local-global navigation in open-world environments.

KLEIYN : A Quadruped Robot with an Active Waist for Both Locomotion and Wall Climbing

Keita Yoneda, Kei Okada

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed the quadruped robot KLEIYN with a waist joint and achieved chimney-like vertical climbing using reinforcement learning

Knowledge Distillation for Semantic Segmentation: A Label Space Unification Approach

Anton Backhaus, Mirko Maehlisch

SegmentationAutonomous DrivingKnowledge DistillationImage

🎯 What it does: This paper proposes a knowledge distillation method for label space unification in semantic segmentation. First, a teacher model is trained using the label system of the source dataset, then pseudo-labels are generated for other datasets with related label spaces using the teacher model, and a student model is trained under the constraints of the mapped label space.

Knowledge-Driven Imitation Learning: Enabling Generalization Across Diverse Conditions

Zhuochen Miao, Cewu Lu

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningDiffusion model

🎯 What it does: Propose a knowledge-driven imitation learning framework that utilizes external structural semantic knowledge to abstract representations of similar objects, introduces semantic keypoint graphs as knowledge templates, and develops a coarse-to-fine template matching algorithm to optimize structural consistency and semantic similarity.

L-SNI: A Language-Driven Semantic Navigation System for Inspection Tasks

Jiawang Ma, Gan Ma

Object DetectionDepth EstimationOptimizationRobotic IntelligenceLarge Language ModelVision Language ModelImageTextMultimodalityPoint Cloud

🎯 What it does: L-SNI is a semantic navigation system that uses LiDAR to construct accurate depth maps, extracts object categories from RGB images to generate semantic maps, and then encodes 3D semantic maps into text for understanding by large language models (LLMs); the LLM decodes natural language commands into inspection primitives to guide low-level planners to execute tasks; simultaneously, a goal cost gradient is proposed to optimize the robot's goal point selection and posture control in semantic maps; after reaching the goal, the scene is described by a vision-language model (VLM), and the LLM simplifies it into a user-friendly response.

L2Calib: SE (3)-Manifold Reinforcement Learning for Robust Extrinsic Calibration with Degenerate Motion Resilience

Baorun Li, Jiajun Lv

Autonomous DrivingOptimizationReinforcement Learning

🎯 What it does: Proposed a reinforcement learning-based SE(3) manifold extrinsic calibration framework, treating calibration as a decision problem and directly optimizing extrinsic parameters to enhance odometry accuracy.

L2COcc: Lightweight Camera-Centric Semantic Scene Completion via Distillation of LiDAR Model

Ruoyu Wang, Xingxing Zuo

Autonomous DrivingComputational EfficiencyKnowledge DistillationTransformerImagePoint Cloud

🎯 What it does: Proposed a lightweight, camera-centric semantic scene completion framework called L2COcc, which supports LiDAR input and utilizes efficient voxel transformers and cross-modal knowledge distillation modules.

L3M+P: Lifelong Planning with Large Language Models

Krish Agarwal, Peter Stone

Robotic IntelligenceTransformerLarge Language ModelMultimodalityRetrieval-Augmented Generation

🎯 What it does: Propose the L3M+P framework, which combines external knowledge graphs with large language models (LLMs) for lifelong planning

Label-Efficient LiDAR Panoptic Segmentation

Ahmet Selim cCanakcci, Abhinav Valada

SegmentationConvolutional Neural NetworkPoint Cloud

🎯 What it does: Propose a label-efficient LiDAR panoramic segmentation method called L3PS, which generates pseudo labels using a 2D network and projects them onto point clouds, then enhances segmentation quality through 3D optimization.

Label-Efficient LiDAR Semantic Segmentation with 2D-3D Vision Transformer Adapters

Julia Hindel, Abhinav Valada

SegmentationAutonomous DrivingRepresentation LearningTransformerMultimodalityPoint Cloud

🎯 What it does: Propose BALViT, a method that utilizes a frozen visual model as an amodal feature encoder to learn a powerful LiDAR encoder, and fuses disparity view and bird's-eye view encoding mechanisms through a 2D-3D adapter;

Label-supervised surgical instrument segmentation using temporal equivariance and semantic continuity

Qiyuan Wang, S. K. Zhou

SegmentationVideoBiomedical DataBenchmark

🎯 What it does: Proposes a label-supervised framework for surgical instrument segmentation, leveraging video temporal features and incorporating temporal equivalence constraints, class-aware semantic continuity constraints, and temporal-enhanced pseudo masks to improve segmentation.

LAMPS: A Novel Robot Generalization Framework for Learning Adaptive Multi-Periodic Skills

Zezhi Liu, Yongchun Fang

Robotic Intelligence

🎯 What it does: Designed and verified a framework called LAMPS for segmenting, learning, and generalizing multi-cycle human skills, enabling robots to learn different motion primitives and execute them in new environments.

Landing-Aware Multi-Drone Routing in Last-Mile Delivery Services

JiHyun Kwon, BaekGyu Kim

Optimization

🎯 What it does: Propose a last-mile delivery route optimization framework under multi-drone landing perception