IROS 2025 Papers — Page 15
IEEE/RSJ International Conference on Intelligent Robots and Systems · 1984 papers
Physics-Based Gas Mapping with Nano Aerial Vehicles: The ADApprox Algorithm
Nicolaj Bösel-Schmid, A. Martinoli
Robotic IntelligencePhysics RelatedOrdinary Differential Equation
🎯 What it does: Developed a physics-based gas distribution mapping algorithm ADApprox, which utilizes local approximation of the transport diffusion equation and learns model parameters from sparse measurements, subsequently predicting gas concentration throughout the entire environment, with its effectiveness validated in nano drone experiments.
Physics-Embedded Neural Networks for sEMG-based Continuous Motion Estimation
Wending Heng, Zhenhong Li
Pose EstimationConvolutional Neural NetworkRecurrent Neural NetworkTime SeriesBiomedical DataPhysics Related
🎯 What it does: Proposed a Physics-Embedded Neural Network (PENN) that integrates musculoskeletal forward dynamics with data-driven residual learning for continuous estimation of movement intent from surface electromyography (sEMG) signals.
Physics-Informed Learning for Human Whole-Body Kinematics Prediction via Sparse IMUs
Cheng Guo, Daniele Pucci
Pose EstimationTime SeriesBiomedical DataPhysics Related
🎯 What it does: Proposed a physics-informed learning framework that predicts full-body kinematics using only 5 IMUs, constructed a network considering spatial features, and incorporated forward and differential kinematics as loss during training, with iterative updates of joint state buffers during inference.
Physics-Informed LSTM for Shape and Contact Force Prediction of a Flexible Surgical Robot*
Feng Ju, Li-bo Ding
Computational EfficiencyRobotic IntelligenceRecurrent Neural NetworkTime SeriesPhysics Related
🎯 What it does: Designed a miniaturized flexible surgical robot with a nested spring structure and proposed a physics-constrained LSTM model for simultaneously predicting the robot's shape and the two-dimensional contact force at the end-effector.
Physics-informed Neural Motion Planning via Domain Decomposition in Large Environments
Yuchen Liu, A. H. Qureshi
Autonomous DrivingOptimizationRobotic IntelligencePhysics Related
🎯 What it does: Proposed Finite Basis Neural Time Fields (FB-NTFields), achieving scalable cost-to-go estimation by computing the distance between start and goal point embeddings in the latent space, and integrating domain decomposition to ensure global spatial coherence in large-scale motion planning.
Physics-informed Neural Time Fields for Prehensile Object Manipulation
Hanwen Ren, A. H. Qureshi
Robotic IntelligenceMultimodalityPhysics RelatedOrdinary Differential Equation
🎯 What it does: Proposes a multimodal physics-informed neural network (PINN) that generates fast object manipulation trajectories by learning the Eikonal equation and achieves reactive replanning of grasping points during manipulation.
Physics-Informed Residual Network for Magnetic Dipole Model Correction and High-Accuracy Localization
Miaozhang Shen, Zixu Wang
Pose EstimationOptimizationConvolutional Neural NetworkPhysics Related
🎯 What it does: Proposed and implemented a physics-informed residual network called PIRNet for correcting simulated magnetic field data from a magnetic dipole model and achieving high-precision magnetic positioning.
PI-WAN: A Physics-Informed Wind-Adaptive Network for Quadrotor Dynamics Prediction in Unknown Environments
Mengyun Wang, Chang Wang
OptimizationRobotic IntelligenceConvolutional Neural NetworkPhysics Related
🎯 What it does: Proposes PI-WAN for predicting quadrotor dynamics in unknown environments and improving closed-loop tracking performance.
PIPE Planner: Pathwise Information Gain with Map Predictions for Indoor Robot Exploration
Seungjae Baek, Jeong hwan Jeon
Robotic IntelligenceImage
🎯 What it does: Proposed the PIPE planner, which integrates cumulative sensor coverage along the path and utilizes map prediction to reduce overestimation of information gain.
PL-VIWO: A Lightweight and Robust Point-Line Monocular Visual Inertial Wheel Odometry
Zhixin Zhang, Pawel Ladosz
Pose EstimationRobotic IntelligenceSimultaneous Localization and MappingImage
🎯 What it does: Proposes a lightweight, robust point-line monocular visual inertial wheel odometry system (VIWO) for localization of ground robots in long-term complex outdoor navigation.
PlaceNet: Obstacle Aware Mobile Manipulator Base Placement through Deep Learning
Alex Navarro, Mitchell Pryor
Robotic IntelligencePoint Cloud
🎯 What it does: Proposes PlaceNet, a deep learning framework for mobile manipulator base placement, addressing the shortcomings of existing methods in obstacle perception and the generalization of learning approaches;
PlanarMesh: Building Compact 3D Meshes from LiDAR using Incremental Adaptive Resolution Reconstruction
Jiahao Wang, Maurice F. Fallon
OptimizationComputational EfficiencyPoint CloudMesh
🎯 What it does: Proposed the PlanarMesh system, achieving online incremental 3D LiDAR map construction and mesh reconstruction, capable of adaptively adjusting mesh resolution to obtain compact yet detail-rich reconstruction results.
Plane detection and ranking via model information optimisation
Daoxin Zhong, Meng Yee Michael Chuah
SegmentationDepth EstimationConvolutional Neural NetworkImage
🎯 What it does: A generic planar detection framework based on model information optimization is proposed, which utilizes probabilistic modeling of depth images and random subsampling to generate multiple candidate models. The information content is calculated using the physical and noise models of depth sensors, identifying the real plane from the model with the least information. Each detected plane is ranked in quality based on the reduction of inlier information.
Planning and Control for Active Morphing Tensegrity Aerial Vehicles in Confined Spaces
Siyuan Hao, Qingkai Yang
OptimizationRobotic Intelligence
🎯 What it does: Proposes a rod-driven tension structure drone utilizing variable rods and cable networks to achieve morphological adaptability and collision resilience, and designs a hierarchical planning framework ensuring the entire drone is confined within an icosahedral space, along with a manifold-based model predictive controller (MPC) for tracking optimized trajectories and compensating for inertial changes during deformation; the framework's navigation capability in confined environments is verified through simulation.
Planning under Uncertainty from Behaviour Trees
Charlie Street, Masoumeh Mansouri
OptimizationRobotic Intelligence
🎯 What it does: Refining behavior trees under uncertainty through planning to improve robot task performance.
Playful DoggyBot: Learning Agile and Precise Quadrupedal Locomotion
Xin Duan, Sören Schwertfeger
Robotic IntelligenceReinforcement Learning
🎯 What it does: Trained and deployed a perception-control separated system based on reinforcement learning, enabling quadruped robots to track and capture airborne objects during high-speed motion.
PLK-Calib: Single-shot and Target-less LiDAR-Camera Extrinsic Calibration using Plücker Lines
Yanyu Zhang, Wei Ren
Pose EstimationAutonomous DrivingImagePoint Cloud
🎯 What it does: Propose two single-shot, target-free LiDAR-camera extrinsic calibration algorithms that utilize line features for calibration.
PlugAndFilter: Architecture Agnostic Booster for Lightweight Registration
Edoardo Malaspina, François Berry
Anomaly DetectionImageVideo
🎯 What it does: Designed the PlugAndFilter framework to enhance multi-modal image registration performance and convert it into a real-time video registration system
PneuChip: A Compact Pneumatic Controller for Large-scale Soft Artificial Muscles
Zheng Wang, Hongying Zhang
Robotic Intelligence
🎯 What it does: Proposed a compact pneumatic controller named PneuChip for large-scale soft robots;
PNEUmorph: a shape-morphing interface comprising a pneumatic membrane constrained by variable-length tendons
Valentina Soana, Helge Wurdemann
🎯 What it does: Proposed and implemented a PNEUmorph deformation interface composed of pneumatic membranes with variable-length tendon constraints, and elaborately discussed its design, geometric characteristics, simulation, fabrication, operation, and evaluation methods.
Point Cloud-Based Control Barrier Functions for Model Predictive Control in Safety-Critical Navigation of Autonomous Mobile Robots
Faduo Liang, Shi-Lu Dai
OptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Propose a real-time motion planning algorithm based on point clouds to enable autonomous mobile robot navigation in safety-critical environments
Policy Learning for Social Robot-Led Physiotherapy
Carl Bettosi, Marta Romeo
Robotic IntelligenceReinforcement LearningBiomedical Data
🎯 What it does: By having 33 expert healthy practitioners simulate patients interacting with social robots, a patient behavior model was constructed to generate movement performance metrics and subjective fatigue scores. Reinforcement learning strategies were trained in a simulated environment, enabling robots to dynamically adjust exercise guidance based on individual tolerance and performance fluctuations, while being applicable to patients at different recovery stages.
Policy Learning from Large Vision-Language Model Feedback Without Reward Modeling
T. Luu, C. Yoo
Robotic IntelligenceReinforcement Learning from Human FeedbackLarge Language ModelVision Language ModelMultimodalityBenchmark
🎯 What it does: Leverage large-scale vision-language models to provide preference labels for offline reinforcement learning, directly training policies from these labels while skipping the construction of explicit reward models.
Pose Estimation of a Cable-Driven Serpentine Manipulator Utilizing Intrinsic Dynamics via Physical Reservoir Computing
Kazutoshi Tanaka, Masashi Hamaya
Pose EstimationRobotic IntelligenceRecurrent Neural Network
🎯 What it does: Developed a 9-degree-of-freedom (DOF), total weight 308g, arm length 545mm cable-driven snake robot arm, and proposed a method for pose estimation utilizing its inherent nonlinear dynamics.
PosePilot: Steering Camera Pose for Generative World Models with Self-supervised Depth
Bu Jin, Hao Zhao
Depth EstimationAutonomous DrivingWorld ModelVideo
🎯 What it does: Propose the PosePilot framework, achieving precise controllability of camera poses in generated world models through self-supervised depth estimation and structured light methods;
Positioning with respect to a cylinder using proximity-based control
John Thomas, François Chaumette
Robotic Intelligence
🎯 What it does: Developed a system with a near-field detection array as the end-effector to complete positioning tasks relative to a cylinder, and designed a classical control law;
Power Balance-Based Recursive Composite Learning Robot Control With Reduced Computational Burden
Tian Shi, Yongping Pan
Robotic Intelligence
🎯 What it does: Proposes a recursive composite learning robot control method based on the power balance model, aiming to enhance parameter convergence speed and tracking control performance.
Precision Autonomous Landing of UAV on High-Speed Vehicles Based on Enhanced Gimbal Stabilization and Smooth Trajectory Generation
Baijian Chen, Kai Hu
OptimizationRobotic Intelligence
🎯 What it does: Proposed a precise autonomous landing system to achieve accurate landing of drones on high-speed moving platforms.
Preference Aligned Diffusion Planner for Quadrupedal Locomotion Control
Xinyi Yuan, Xuelong Li
Robotic IntelligenceDiffusion model
🎯 What it does: Proposed a two-stage learning framework, where the offline phase learns the joint distribution of state-action sequences from expert datasets, the online phase interacts in simulation environments to enhance diversity, and introduced a weak preference annotation method, ultimately achieving zero-shot transfer on the real Unitree Go1 robot.
Preferenced Oracle Guided Multi-mode Policies for Dynamic Bipedal Loco-Manipulation
Prashanth Ravichandar, Quan Nguyen
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes Preferenced Oracle Guided Multi-mode Policies (OGMP), achieving multi-mode dynamic locomotion-manipulation via a single policy;
PrefMMT: Modeling Human Preferences in Preference-based Reinforcement Learning with Multimodal Transformers
Dezhong Zhao, Guohua Chen
Reinforcement Learning from Human FeedbackTransformerReinforcement LearningMultimodality
🎯 What it does: Propose a multi-modal sequence modeling method that separates state and action modalities and designs a multi-modal Transformer network called PrefMMT for more accurate modeling of human preferences.
Prescribed-Time Safe Pursuit Control with Dynamic Obstacle and Occlusion Avoidance
Zheng Li, Qinglei Hu
Autonomous DrivingSafty and Privacy
🎯 What it does: A safety-critical tracking control method is proposed to ensure dynamic obstacles remain outside the camera's field of view while avoiding collisions between the tracking vehicle and obstacles. A real-time occlusion detection function is developed, and motion constraints are systematically integrated into a hybrid framework combining artificial potential field (APF) and observer control strategies. A prescribed-time controller (PTC) based on time-scale transformation and a simplified structured prescribed-time linear extended state observer (PTESO) are proposed to rapidly and accurately estimate unknown environmental disturbances and nonlinear terms. The effectiveness of the method is finally validated in simulations of simplified physical scenarios.
Proactive Conflict Area Prediction for Boosting Search-Based Multi-Agent Pathfinding
Youngjoon Kwon, Kyungjae Lee
Optimization
🎯 What it does: Proposes a Proactive Conflict-Aware Prediction (PCAP) method, which improves traditional Conflict-Based Search (CBS) by predicting conflict-prone areas, thereby reducing invalid expansions of the constraint tree.
Probabilistic Collision Risk Estimation for Pedestrian Navigation
Amine Tourki, Alexandre Alahi
Autonomous DrivingSafty and Privacy
🎯 What it does: Integrate the probability collision risk model, which has already been used in autonomous driving and advanced driver assistance systems, into assistive devices for visually impaired individuals, and experimentally evaluate its performance in alarm functions.
Probabilistic Motion Model Learning for Tendon Actuated Continuum Robots with Backlash
Mahdi Chaari, F. Nageotte
Robotic Intelligence
🎯 What it does: A probabilistic motion model is proposed for tendon-driven continuum robots, addressing the nonlinearity in transmission caused by backlash and cable friction, with Bayesian parameter estimation used to learn parameter distributions and achieve uncertainty propagation.
Project Yukionna: Fabrication of Ice-Based Robotic Components via Formative Methods
Kunlun Wu, Vlasov Sergey
Robotic Intelligence
🎯 What it does: Develop and verify the ice-based forming method (IFM), assess the feasibility of forming manufacturing (FM), subtractive manufacturing (SM), and additive manufacturing (AM) in ice-based robotic components, and complete most of the production processes and experimental validation under full outdoor conditions.
Prototypes, Mathematical Modeling and Motion Analysis of Heptagonal Passive Rotating Locomotion Robots with Elastic Elements Arranged on Diagonal Lines
Fumihiko Asano, Isao T. Tokuda
Robotic Intelligence
🎯 What it does: Proposed and experimentally verified two new heptagonal passive rotational motion robot models, moving the elastic elements from the rotating joints to the diagonals to achieve internal tissue flexible simulation.
Pseudo Depth Meets Gaussian: A Feed-forward RGB SLAM Baseline
Linqing Zhao, Jiwen Lu
Pose EstimationDepth EstimationGaussian SplattingSimultaneous Localization and MappingOptical FlowImageVideo
🎯 What it does: Propose an online 3D reconstruction method based on 3D Gaussian mapping, which integrates a feedforward recurrent prediction module to directly estimate camera pose from optical flow, and introduces a local graph rendering technique to enhance the robustness of pose prediction.
Pursuit-Evasion for Car-like Robots with Sensor Constraints
Burak M. Gonultas, Volkan Isler
Autonomous DrivingRobotic IntelligenceReinforcement Learning
🎯 What it does: Studied pursuit-evasion games with vehicle dynamics and perception constraints, and proposed a learning method that encodes historical information into belief states to generate agent actions.
PyRoki: A Modular Toolkit for Robot Kinematic Optimization
Chung Min Kim, Angjoo Kanazawa
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Developed PyRoki, an extensible, device-agnostic robotic motion optimization toolkit that provides a unified kinematic variable and cost specification interface along with an efficient nonlinear least squares optimizer.
Q-Learning-based Optimal Force-Tracking Control of Grinding Robots in Uncertain Environments
Rui Yang, Qinglei Hu
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a dual-loop force tracking control framework based on Q-learning for robotic grinding tasks, achieving precise force tracking and minimal overshoot in uncertain environments.
QBIT: Quality-Aware Cloud-Based Benchmarking for Robotic Insertion Tasks
Constantin Schempp, Björn Hein
Robotic IntelligenceBenchmark
🎯 What it does: Proposed and implemented QBIT, a quality-aware cloud benchmark framework for robot insertion tasks, adding metrics such as force energy, force smoothness, and completion time. Evaluated geometry-based, force-based, and learning-based insertion methods through randomized contact parameters in MuJoCo, large-scale experiments on Kubernetes, microservices architecture, and containerized ROS2, with comparisons conducted in both simulated and real-world environments.
QLIO: Quantized LiDAR-Inertial Odometry
Boyang Lou, Enwen Hu
Pose EstimationCompressionOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed QLIO, a multi-processor distributed quantized LiDAR-inertial odometry framework, reducing computational load and bandwidth consumption while maintaining localization accuracy.
Quadrotor Morpho-Transition: Learning vs Model-Based Control Strategies
Ioannis Mandralis, Morteza Gharib
Robotic IntelligenceReinforcement Learning
🎯 What it does: Trained and implemented an end-to-end reinforcement learning (RL) controller for drones during the aerial-to-ground morpho-transition process, successfully migrated it to a hardware platform, and compared it with traditional model predictive control (MPC) baseline methods.
Quality-Driven Adaptive Control Framework for Robotic Ultrasound Imaging of Vascular Anatomies
Bo Wang, E. Momi
SegmentationRobotic IntelligenceConvolutional Neural NetworkBiomedical DataUltrasound
🎯 What it does: Proposed a quality-driven adaptive control framework for robotic-assisted vascular anatomy ultrasound imaging, incorporating probabilistic image quality assessment metrics, mapping between image segmentation networks and robotic control variables, and quality-driven probe control strategies.
Quantifying and Modeling Driving Styles in Trajectory Forecasting
L. Zheng, Ming C. Lin
Autonomous DrivingTime SeriesSequential
🎯 What it does: Analyze existing real-world driving trajectory datasets and decompose related work from the perspective of driving style, exploring the role of driving style in trajectory prediction
Quasi-God Object and Geodesically Restricted 6-DOF Haptic Forces For Compliant Constraints and Low Frequency Simulation
Ignacio Montesino, Alberto Jardón Huete
Robotic Intelligence
🎯 What it does: Propose a novel method that combines a quasi-God object with a penalty method to achieve 6-degree-of-freedom (DoF) tactile force generation for collaborative robots in a low-frequency commercial game engine environment, ensuring collision compliance and safety through relaxed quasi-God object simulation and geometric constraints.
Quaternion Approximate Networks for Enhanced Image Classification and Oriented Object Detection
Bryce Grant, Peng Wang
ClassificationObject DetectionConvolutional Neural NetworkImage
🎯 What it does: Proposed the Quaternion Approximate Networks (QUAN) framework, utilizing quaternion algebra to achieve rotation equivariant image classification and oriented object detection, and approximating quaternion convolutions through Hamilton decomposition; simultaneously introducing independent quaternion batch normalization and extending quaternion operations to spatial attention mechanisms.
QueryAdapter: Rapid Adaptation of Vision-Language Models in Response to Natural Language Queries
N. H. Chapman, Chris Lehnert
Object DetectionRetrievalDomain AdaptationComputational EfficiencyRepresentation LearningPrompt EngineeringVision Language ModelImageTextPoint Cloud
🎯 What it does: Propose QueryAdapter, which can quickly adapt pre-trained vision-language models based on natural language queries within minutes.
QuietPaw: Learning Quadrupedal Locomotion with Versatile Noise Preference Alignment
Yuyou Zhang, Ding Zhao
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes the QuietPaw framework and achieves sustainable regulation of foot noise in quadruped robot motion control while maintaining mobility through Conditional Noise-Constrained Policy (CNCP).
R-Tac0: A Rounded High-Frequency Transferable Monochrome Vision-based Tactile Sensor for Shape Reconstruction
Wanlin Li, Hangxin Liu
Pose EstimationDepth EstimationRobotic IntelligenceConvolutional Neural NetworkImage
🎯 What it does: Developed a low-cost, high-resolution, high-frequency circular visual tactile sensor named R-Tac0;
R2LDM: An Efficient 4D Radar Super-Resolution Framework Leveraging Diffusion Model
Boyuan Zheng, Lu Xiong
Super ResolutionDiffusion modelPoint Cloud
🎯 What it does: Generate dense and accurate radar point clouds from original 4D radar data using paired LiDAR point clouds as guidance, producing LiDAR-like point clouds
R2Nav: Robust, Real-time Test Time Adaptation for Robot Assisted Endoluminal Navigation
Junyang Wu, Guang-Zhong Yang
Domain AdaptationRobotic Intelligence
🎯 What it does: Proposed R2Nav, a robust real-time test-time adaptation method for robot-assisted endoscopic navigation.
RA-DP: Rapid Adaptive Diffusion Policy for Training-Free High-frequency Robotics Replanning
Xiaohan Ye, Amir Rasouli
Robotic IntelligenceDiffusion model
🎯 What it does: Proposes RA-DP, a training-free high-speed re-planning diffusion strategy framework that achieves instant re-planning at each denoising step by integrating guidance signals during diffusion sampling and employing an action queue mechanism, adapting to dynamic environments
RA-NeRF: Robust Neural Radiance Field Reconstruction with Accurate Camera Pose Estimation under Complex Trajectories
Qingsong Yan, Fei Deng
Pose EstimationNeural Radiance FieldGaussian Splatting
🎯 What it does: Proposed the RA-NeRF method, achieving robust neural radiance field reconstruction and precise camera pose estimation under complex camera trajectories.
Radar-Based NLoS Pedestrian Localization for Darting-Out Scenarios Near Parked Vehicles with Camera-Assisted Point Cloud Interpretation
Hee-Yeun Kim, Seong-Woo Kim
Object DetectionSegmentationDepth EstimationConvolutional Neural NetworkSimultaneous Localization and MappingImageMultimodalityPoint Cloud
🎯 What it does: Proposes a Non-Line-of-Sight (NLoS) pedestrian localization framework that integrates monocular camera images with 2D radar point clouds, first detecting parked vehicles and estimating depth through image segmentation, then refining spatial inference using radar point clouds.
RAG-6DPose: Retrieval-Augmented 6D Pose Estimation via Leveraging CAD as Knowledge Base
Kuanning Wang, X. Xue
Pose EstimationMultimodalityRetrieval-Augmented Generation
🎯 What it does: Proposes RAG-6DPose, a retrieval-enhanced 6D pose estimation method that leverages a CAD model knowledge base.
RaGNNarok: A Light-Weight Graph Neural Network for Enhancing Radar Point Clouds on Unmanned Ground Vehicles
David Hunt, Miroslav Pajic
Autonomous DrivingRobotic IntelligenceGraph Neural NetworkPoint Cloud
🎯 What it does: Proposes RaGNNarok, a real-time, lightweight, and generalizable graph neural network framework for enhancing millimeter-wave radar point clouds to improve the performance of low-cost indoor mobile robots.
RAILGUN: A Unified Convolutional Policy for Multi-Agent Path Finding Across Different Environments and Tasks
Yimin Tang, Sven Koenig
OptimizationConvolutional Neural NetworkImage
🎯 What it does: Developed RAILGUN—the first centralized learning strategy for multi-agent path finding, using CNN to encode maps and generalize across different maps and any number of agents.
RainforestDepth: Monocular Depth Estimation Targeting Rainforest Environments
Sri Sai Anirudh Tangellapalli, Brittany A. Duncan
Depth EstimationImage
🎯 What it does: Proposed a monocular depth estimation model specifically tailored for rainforest environments, combined with a new dataset integrating synthetic and real images.
RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with Retrieval-Augmented Learning
Jiacheng Zuo, C. Xue
Domain AdaptationAutonomous DrivingRetrieval-Augmented Generation
🎯 What it does: Proposed and implemented Retrieval-Augmented Learning for Autonomous Driving (RALAD), which bridges the gap between real and simulated domains by retrieving the most similar real and simulated scenarios, performing cross-scenario feature fusion, and freezing feature extraction while fine-tuning on fused features.
Rapid and Simultaneous Visual-based Estimation of Kinematic and Hand-eye Parameters of Industrial Mobile Manipulators
S. Mutti, Anna Valente
Robotic IntelligenceImage
🎯 What it does: Utilizing a standard 2D camera system, employs the Unscented Kalman Filter (UKF) combined with uncertainty propagation of robotic kinematic parameters to real-time simultaneously estimate the kinematic parameters and eye-hand transformation matrix of an industrial mobile manipulator, and iteratively improve parameter estimation;
Rapid Flight Trajectory Planning for Autonomous Terrain Avoidance via Generative Learning
A. T. Çetin, Emre Koyuncu
OptimizationRobotic IntelligenceGenerative Adversarial NetworkPoint Cloud
🎯 What it does: Propose a parallel autonomous system that utilizes a sampling-based motion planner combined with generative adversarial learning and differential flatness techniques to generate terrain-avoidance motion trajectories for aircraft in real-time against the background, and immediately take over control when necessary.
RAVES-Calib: Robust, Accurate and Versatile Extrinsic Self Calibration Using Optimal Geometric Features
Haoxin Zhang, Wen Yao
Autonomous DrivingImagePoint Cloud
🎯 What it does: Propose a user-friendly LiDAR-camera self-calibration toolbox compatible with various sensors, which can complete self-calibration in a target-free environment using only a pair of laser points and camera images.
Ray Visual Odometry
Fanqi Xu, N. Trigoni
Pose EstimationNeural Radiance Field
🎯 What it does: Proposes modeling visual odometry as a dense distributed representation called RayVO, which is trained using three specialized loss functions to simultaneously predict camera intrinsic and extrinsic parameters.
RayFronts: Open-Set Semantic Ray Frontiers for Online Scene Understanding and Exploration
Omar Alama, Sebastian A. Scherer
SegmentationComputational EfficiencyRepresentation Learning
🎯 What it does: Propose the RayFronts unified representation method, achieving efficient semantic mapping for both in-range and super-distance scenarios.
RCGNet: RGB-based Category-Level 6D Object Pose Estimation with Geometric Guidance
Sheng Yu, Yuanqing Xia
Pose EstimationTransformerImage
🎯 What it does: Proposes a category-level object pose estimation method that relies solely on RGB images, utilizing a Transformer network to predict and fuse object geometric features, employing a geometric feature-guided algorithm to ensure the accuracy of geometric information, and finally using RANSAC-PnP to compute the pose.
RDMM: Enhancing Household Robotics with On-Device Contextual Memory and Decision Making
Shady Nasrat, Seung-Joon Yi
Robotic IntelligenceTransformerLarge Language ModelMultimodality
🎯 What it does: Proposes the RDMM framework, which enhances robot autonomy by leveraging large language models for domain-specific robot decision-making, integrating agent-specific knowledge representation, visual perception, and real-time speech recognition;
RDN: An Efficient Denoising Network for 4D Radar Point Clouds
Ningyuan Huang, Zheng Fang
RestorationAutonomous DrivingPoint Cloud
🎯 What it does: Proposed a denoising network RDN for 4D radar point clouds, including the feature similarity-based farthest point sampling module (FS-FPS), virtual feature point prediction module (VFP), and iterative upsampling module (IUS), to enhance the quality of point clouds for robot perception and autonomous driving.
REACT: Real-time Efficient Attribute Clustering and Transfer for Updatable 3D Scene Graph
Phuoc Nguyen, Ville Kyrki
Computational EfficiencyRepresentation LearningContrastive LearningGraph
🎯 What it does: Developed the REACT framework, capable of real-time and efficient clustering and transfer of object attributes in 3D scene graphs to achieve repositioning of object nodes.
Reactive 3D Motion Planning in Dynamic Environments Using Efficient Model Predictive Control via Circular Fields *
Fabrice Zeug, M. A. Müller
Autonomous DrivingOptimizationRobotic Intelligence
🎯 What it does: Proposed an online global reactive motion planner based on circular fields, combining reactive control and MPC.
Reactive Model Predictive Contouring Control for Robot Manipulators
Junheon Yoon, Jaeheung Park
OptimizationRobotic Intelligence
🎯 What it does: Propose a robot path tracking framework based on Reactive Model Predictive Contouring Control (RMPCC), capable of avoiding obstacles, singularities, and self-collision at a frequency of 100 Hz.
Reactive Temporal Logic Planning for Safe Human-Robot Interaction
Xiangcheng Liu, Zhen Kan
Safty and PrivacyRobotic Intelligence
🎯 What it does: Proposed a reactive task and motion planning framework based on Linear Temporal Logic (LTL) to achieve safe and efficient human-robot interaction experiments
Real-Time 3D Guidewire Reconstruction from Intraoperative DSA Images for Robot-Assisted Endovascular Interventions
Tianliang Yao, P. Qi
Robotic IntelligenceImageMultimodalityBiomedical DataComputed Tomography
🎯 What it does: Developed a real-time 3D catheter reconstruction framework that integrates preoperative CTA and intraoperative 2D DSA images.
Real-Time Consistent Monocular Depth Recovery System for Dynamic Environments
Gan Huang, Guofeng Zhang
Depth EstimationTransformerSimultaneous Localization and MappingImage
🎯 What it does: Proposed a real-time consistent monocular depth recovery system by fusing ORB-SLAM3's sparse depth initialization, a ViT-based depth completion network, a motion segmentation module, and a dual-weight fusion module, achieving the generation of globally scale-consistent, spatiotemporally consistent, and dense depth maps.
Real-time Distributed Force Sensing-Based Position Feedback Control for Fiber-Driven Miniaturized Continuum Robots
Jingyuan Xia, A. Gao
Robotic Intelligence
🎯 What it does: Reconstructing shape based on fiber optic distributed force sensing, and utilizing this shape to achieve real-time position closed-loop control for fiber-driven continuum robots;
Real-Time Guaranteed Monitoring for a Drone Using Interval Analysis and Signal Temporal Logic
Antoine Besset, J. Tillet
Autonomous Driving
🎯 What it does: Proposed a model-based real-time safety monitoring method that uses interval analysis and signal temporal logic to monitor UAV trajectories, and implemented real-time monitoring and adjustment on the ROS platform.
Real-Time Incremental Mapping and Degeneration-Aware Localization for Multi-Floor Parking Lots Based on IPM Image
Youyang Feng, Dingsheng Luo
Autonomous DrivingOptimizationSimultaneous Localization and MappingImage
🎯 What it does: Proposed a real-time incremental mapping and degradation-aware localization system applicable to multi-level parking lots.
Real-Time Initialization of Unknown Anchors for UWB-aided Navigation
Giulio Delama, Stephan Weiss
Anomaly DetectionAutonomous DrivingOptimization
🎯 What it does: Propose a real-time framework for initializing unknown UWB anchors, which can automatically detect and calibrate previously unknown anchors during runtime, eliminating the need for manual setup.
Real-time Iteration Scheme for Diffusion Policy
Yufei Duan, Danica Kragic
Computational EfficiencyDiffusion model
🎯 What it does: Proposes a diffusion strategy inference acceleration method based on a real-time iteration (RTI) scheme, utilizing the solution from the previous step as the initial guess for the next step, and applying scaling processing to discrete actions.
Real-Time Manipulation Action Recognition with a Factorized Graph Sequence Encoder
Enes Erdogan, Sanem Sariel
RecognitionComputational EfficiencyRepresentation LearningGraph Neural NetworkVideo
🎯 What it does: Proposed a Factorized Graph Sequence Encoder network for real-time recognition of human manipulation actions, and introduced the Hand Pooling operation to enhance the focus of graph layer embeddings.
Real-Time Occupancy Grid Mapping Using RMM on Large-scale and Unstructured Environments
Xingyu Li, Zhaotong Tan
Computational EfficiencyRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: A lightweight mapping framework based on the Random Mapping Method (RMM) is adopted to achieve real-time occupancy grid mapping for drones in large-scale unstructured environments.
Real-Time Optimization-Based Quadrotor Trajectory Generation with Kinodynamic Constraints in Unknown Environments
Pinhui Zhao, Yuqing He
OptimizationRobotic Intelligence
🎯 What it does: Propose a real-time optimized quadrotor trajectory generation method that integrates dynamic constraints into the trajectory search and optimization stages;
Real-time Photorealistic Mapping for Situational Awareness in Robot Teleoperation
I. Page, P. Wieber
Computational EfficiencyRobotic IntelligenceGaussian SplattingSimultaneous Localization and Mapping
🎯 What it does: Propose a modular, efficient GPU-based integrated solution that combines Gaussian Splatting SLAM with an existing online map-based remote operating system to enhance remote operation efficiency in unknown environments.
Real-Time Position-Based Deformable Human Body Dynamics for Disaster Rescue Simulation: A Stress-Driven Approach using a Practical Neo-Hookean Constraint
Xu Wang, Atsushi Konno
Computational EfficiencyMesh
🎯 What it does: Developed a fully GPU-parallel, pressure-driven real-time human deformation dynamics simulation framework specifically for disaster rescue simulations.
Real-Time Reinforcement Learning for Dynamic Tasks with a Parallel Soft Robot
James Avtges, Todd D. Murphey
Computational EfficiencyRobotic IntelligenceReinforcement LearningDiffusion model
🎯 What it does: Achieve dynamic balance control of a soft Stewart platform in a single deployment using reinforcement learning, expanding the balance neighborhood through curriculum learning strategies and maintaining balance even when some actuators fail.
Real-time Spatial-temporal Traversability Assessment via Feature-based Sparse Gaussian Process
Senming Tan, Yanjun Cao
Autonomous DrivingComputational EfficiencyPoint Cloud
🎯 What it does: Propose a real-time spatiotemporal traversability assessment method, which utilizes sparse Gaussian processes to extract point cloud geometric features and construct a high-resolution local traversability map, then designs a spatiotemporal Bayesian Gaussian kernel to infer real-time traversability scores.
Real-time Whole-body Motion Planning Based on Optimized NMPC in Static and Dynamic Environments for Mobile Manipulator
Wei Wu, Guiyang Xin
OptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Proposes an integrated framework for environment perception, real-time planning, and control optimization, which includes map construction by fusing ESDF with clustered point clouds, rapid generation of 6-DOF guided point sequences, and a full-body motion controller using optimized nonlinear model predictive control (NMPC), implemented on a mobile manipulator with an Ackerman chassis and validated through simulation and field experiments;
Real-World Offline Reinforcement Learning from Vision Language Model Feedback
Sreyas Venkataraman, David Held
Robotic IntelligenceReinforcement Learning from Human FeedbackTransformerReinforcement LearningVision Language ModelMultimodality
🎯 What it does: Built a system that automatically generates reward labels for an offline dataset using a vision-language model and preference feedback from task descriptions, and learns a policy based on this using offline reinforcement learning (Implicit Q Learning), verifying its effectiveness in real robot-assisted dressing tasks and simulated object manipulation tasks.
REALMS2 - Resilient Exploration And Lunar Mapping System 2 – A Comprehensive Approach
Dave van der Meer, M. Olivares-Méndez
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Propose the REALMS2 multi-robot system framework for planetary exploration and mapping, based on ROS 2, vSLAM, mesh networks, and a single graphical interface, achieving scenario testing and approximately 60% area mapping.
ReBot: Scaling Robot Learning with Real-to-Sim-to-Real Robotic Video Synthesis
Yu Fang, Mingyu Ding
Data SynthesisDomain AdaptationRobotic IntelligenceVision-Language-Action ModelVideo
🎯 What it does: ReBot expands real robot data and helps VLA models adapt to target domains by replaying real robot trajectories into simulation environments, then synthesizing physically realistic and temporally coherent robot videos by combining simulated motion with inpainted real backgrounds.
Recognizing and Generating Novel Emotional Behaviors on Two Robotic Platforms
Rista Baral, C. Kennington
GenerationRobotic IntelligenceTransformerLarge Language ModelGenerative Adversarial Network
🎯 What it does: Proposed an adversarial training scheme that utilizes an emotion discriminator and novelty loss to optimize the emotion behavior generation model, and validated its effectiveness through experiments and human evaluation.
Recognizing Skeleton-Based Actions As Points
Baiqiao Yin, Mengyuan Liu
RecognitionPoint Cloud
🎯 What it does: Propose the Skeleton2Point network, treating skeleton action recognition as point clouds and enhancing feature expression through the Information Transformation Module (ITM) and Cluster Dispatch Interaction Module (CDI).
Recommendation Navigation Based on User Information Using VLM
Daewon Kwak, Donghan Kim
Recommendation SystemLarge Language ModelVision Language ModelImageText
🎯 What it does: Proposed a recommendation-based path planning system that utilizes VLM and LLM to parse user intent, providing personalized navigation and guidance services.
RECON: Reducing Causal Confusion with Human-Placed Markers
Robert Ramirez Sanchez, Dylan P. Losey
Robotic IntelligenceReinforcement Learning from Human FeedbackSupervised Fine-Tuning
🎯 What it does: Propose a framework named RECON, which utilizes lightweight beacons placed by humans on critical objects before demonstrations, and subsequently trains task-related state embeddings associated with beacon readings through offline beacon data, thereby reducing causal confusion and decreasing the required number of demonstrations in imitation learning.
RecoveryChaining: Learning Local Recovery Policies for Robust Manipulation
Shivam Vats, Diego Romeres
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a hierarchical reinforcement learning-based recovery strategy to detect failures and restore the robot to a usable state when executing model-driven planning and control.
Reducing Redundancy in VSLAM: VLMs-driven Keyframe Selection using Multi-dimensional Semantic Information
Xiang Huo, Weinan Chen
Autonomous DrivingTransformerVision Language ModelSimultaneous Localization and MappingMultimodality
🎯 What it does: Proposed a keyframe selection method that utilizes a vision-language model to extract multi-dimensional semantic features and combines them with Bayesian online change point detection.
Reducing Scene Graph Generation Parameters Towards UAV Understanding of Structured Environments
Xudong Li, Lizhen Wu
Autonomous DrivingComputational EfficiencyRepresentation LearningTransformer
🎯 What it does: Propose a streamlined scene graph generation model to reduce the number of parameters in UAV applications, directly using subject-object query pairs to predict triples;
Refer and Grasp: Vision-Language Guided Continuous Dexterous Grasping
Yayu Huang, Peng Wang
Data SynthesisRobotic IntelligenceVision Language ModelVision-Language-Action ModelMultimodality
🎯 What it does: Propose an automated dataset generation engine and a visual language-guided continuous dexterous grasping framework, and validate its effectiveness on real robots.
Reference-Steering via Data-Driven Predictive Control for Hyper-Accurate Robotic Flying-Hopping Locomotion
Yicheng Zeng, Xiaobin Xiong
Robotic Intelligence
🎯 What it does: A reference regulation method is proposed, utilizing data-driven input-output models and predictive control for online adjustment of reference trajectories, achieving ultra-precise jumping and flight behaviors on the robot PogoX;