arXivSub Start free trial

ICRA 2024 Papers — Page 8

IEEE International Conference on Robotics and Automation · 1760 papers

HAGrasp: Hybrid Action Grasp Control in Cluttered Scenes using Deep Reinforcement Learning

Kai-Tai Song, Hsiang-Hsi Chen

Robotic IntelligenceReinforcement LearningPoint Cloud

🎯 What it does: Proposed and implemented a closed-loop hybrid action grasping control system that integrates grasp pose estimation, end-effector pose evaluation, and robotic arm motion planning into a single model.

Hamiltonian Dynamics Learning from Point Cloud Observations for Nonholonomic Mobile Robot Control

Abdullah Altawaitan, Nikolay Atanasov

Robotic IntelligencePoint CloudOrdinary Differential Equation

🎯 What it does: Developed a method to learn robot dynamics directly from point cloud observations, utilizing neural ODEs with embedded Hamiltonian structures to train dynamics models, and designed an energy-regulating model-based tracking controller based on this model, demonstrating the effectiveness of dynamics learning and tracking control on a real non-homogeneous wheeled robot.

HandNeRF: Learning to Reconstruct Hand-Object Interaction Scene from a Single RGB Image

Hongsuk Choi, Hyunsoo Park

GenerationNeural Radiance FieldImage

🎯 What it does: Learn hand-object interaction priors from a single RGB image, leverage hand shape to constrain hand-object geometry, and perform 3D reconstruction of hand-object scenes using the implicit function HandNeRF.

HandyPriors: Physically Consistent Perception of Hand-Object Interactions with Differentiable Priors

Shutong Zhang, Animesh Garg

Pose EstimationOptimizationRobotic IntelligencePhysics Related

🎯 What it does: Propose a unified hand-object interaction pose estimation pipeline called HANDYPRIORS, which utilizes differentiable physics and rendering for pose inference;

HAPFI: History-Aware Planning based on Fused Information

Sujin Jeon, Byoung-Tak Zhang

Robotic IntelligenceVision-Language-Action ModelImageTextMultimodality

🎯 What it does: Proposes HAPFI, a planning method for embedded instruction following using a history-aware multimodal information fusion approach.

Haptic-Assisted Collaborative Robot Framework for Improved Situational Awareness in Skull Base Surgery

Hisashi Ishida, Russell H. Taylor

Safty and PrivacyRobotic IntelligenceBiomedical Data

🎯 What it does: Developed and tested a collaborative control robot system equipped with haptic assistive modes for drilling in cranial base surgery, verifying its safety and efficiency.

Hard Shell, Soft Core: Binary Actuators for Deep-Sea Applications

Cora Maria Sourkounis, Annika Raatz

Robotic Intelligence

🎯 What it does: Proposed and verified a binary drive system integrating soft actuation with rigid bistable mechanisms for deep-sea suction cup samplers to reduce costs.

Harnessing the Differential Flatness of Monocopter Dynamics for the Purpose of Trajectory Tracking in a Stable Invertible Coaxial Actuated ROtorcraft (SICARO)

Emmanuel Tang, S. Foong

Robotic IntelligencePhysics Related

🎯 What it does: This paper proves that the dynamics of Monocopter are differentially flat, and utilizes this property to compute feedforward terms for trajectory tracking; taking SICARO as an example, the feedforward terms are integrated into a cascaded nonlinear controller, and the effectiveness of the method is validated in flight experiments on both wings.

Harnessing the Synergy between Pushing, Grasping, and Throwing to Enhance Object Manipulation in Cluttered Scenarios

H. Kasaei, M. Kasaei

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed a coordinated control process for robots to rearrange objects in cluttered scenes through pushing, grasping, and throwing.

HASHI: Highly Adaptable Seafood Handling Instrument for Manipulation in Industrial Settings

Austin Allison, T. Padır

Robotic Intelligence

🎯 What it does: A novel robotic end-effector named HASHI was proposed, which employs chopstick-like appendages to achieve precise and flexible grasping. A 6-axis torque sensor is embedded for food localization and posture control. Its kinematic model was validated, and its diversity was evaluated through real and simulated food grasping experiments.

Hearing Touch: Audio-Visual Pretraining for Contact-Rich Manipulation

Jared Mejia, Abhinav Gupta

Representation LearningRobotic IntelligenceVideoMultimodalityAudio

🎯 What it does: Utilizing contact microphones as tactile sensors and obtaining robust multisensory representations through large-scale audio-visual pretraining to enhance robot manipulation performance.

HEGN: Hierarchical Equivariant Graph Neural Network for 9DoF Point Cloud Registration

Adam Misik, Eckehard G. Steinbach

Pose EstimationGraph Neural NetworkPoint Cloud

🎯 What it does: Propose a HEGN network for 9DoF point cloud registration, utilizing equivariance to achieve transformation estimation without relying on corresponding points.

Helical Control in Latent Space: Enhancing Robotic Craniotomy Precision in Uncertain Environments

Yuanyuan Jia, T. Taniguchi

Robotic Intelligence

🎯 What it does: Propose a dual-stage transfer learning framework for helical control in robotic craniotomy surgery to enhance precision.

HERO-SLAM: Hybrid Enhanced Robust Optimization of Neural SLAM

Zhe Xin, Chenming Wu

Neural Radiance FieldSimultaneous Localization and Mapping

🎯 What it does: Propose HERO-SLAM, a hybrid enhanced robust optimization neural SLAM method.

Heuristic-based Incremental Probabilistic Roadmap for Efficient UAV Exploration in Dynamic Environments

Zhefan Xu, Kenji Shimada

Robotic Intelligence

🎯 What it does: Proposes a heuristic incremental probabilistic roadmap (HIRE) planner for efficient and safe exploration of unmanned aerial vehicles in dynamic environments, employing an incremental sampling strategy based on heuristic frontier detection, and equipped with a dynamic module to update the roadmap in real-time and avoid dynamic obstacles.

HHGNN: Heterogeneous Hypergraph Neural Network for Traffic Agents Trajectory Prediction in Grouping Scenarios

Hetian Guo, Xuan Song

Autonomous DrivingGraph Neural NetworkTime Series

🎯 What it does: Proposed the HHGNN model, which encodes inter-group interactions of traffic participants using heterogeneous hypergraphs, and captures intra- and inter-group interactions of heterogeneous agents through a type-aware two-layer hypergraph message passing module;

HIC-YOLOv5: Improved YOLOv5 For Small Object Detection

Shiyi Tang, Shu Zhang

Object DetectionConvolutional Neural NetworkImage

🎯 What it does: Proposed the HIC-YOLOv5 model, improving YOLOv5 to enhance small object detection performance

Hierarchical Deep Learning for Intention Estimation of Teleoperation Manipulation in Assembly Tasks

Mingyu Cai, Songpo Li

Robotic Intelligence

🎯 What it does: Propose a hierarchical deep learning framework for inferring user intent in teleoperated assembly tasks.

Hierarchical Experience-informed Navigation for Multi-modal Quadrupedal Rebar Grid Traversal

Max Asselmeier, Ye Zhao

Robotic Intelligence

🎯 What it does: A hierarchical, experience-based multimodal contact planning framework was developed for quadruped robots to achieve agile locomotion in constrained rebar environments, integrated with global torso path guidance.

Hierarchical Human-to-Robot Imitation Learning for Long-Horizon Tasks via Cross-Domain Skill Alignment

Zhenyang Lin, Zhiyong Liu

Domain AdaptationRobotic IntelligenceVideo

🎯 What it does: Proposed a hierarchical human-to-robot imitation learning framework called H2RIL, which achieves zero-shot task generalization across different environments by learning cross-domain sensorimotor skill mappings through human videos.

Hierarchical Learned Risk-Aware Planning Framework for Human Driving Modeling

Nathan Ludlow, John Dolan

Autonomous DrivingRecurrent Neural Network

🎯 What it does: Proposed a learning-based human driving behavior modeling method based on a hierarchical foresight and risk-aware estimation framework.

Hierarchical Meta-learning-based Adaptive Controller

Fengze Xie, Soon-Jo Chung

Meta Learning

🎯 What it does: Designed a Hierarchical Meta-Learning based Adaptive Controller (HMAC) to achieve fast and accurate online adaptation to multi-source disturbances

Hierarchical Optimization-based Control for Whole-body Loco-manipulation of Heavy Objects

Alberto Rigo, Quan Nguyen

OptimizationRobotic Intelligence

🎯 What it does: Proposed a hierarchical optimization-based whole-body coordination control framework for legged robots to move and manipulate heavy objects.

Hierarchical Planning for Long-Horizon Multi-Agent Collective Construction

Shambhavi Singh, H. Choset

OptimizationRobotic Intelligence

🎯 What it does: Propose a hierarchical planning method that first uses A* search to determine the placement and removal order of blocks, then identifies sequential constraints, and finally plans collision-free paths for multiple robots to construct 3D target structures.

Hierarchical Point Attention for Indoor 3D Object Detection

Manli Shu, Ran Xu

Object DetectionTransformerPoint Cloud

🎯 What it does: Proposed two hierarchical attention modules (MS-A and Local-A) to enhance point cloud Transformers for indoor 3D object detection.

High Precision Paint Deposition Modeling Considering Variable Posture of Spray Painting Robot

Genichiro Tanaka, Hiroyasu Iwata

Robotic IntelligencePhysics Related

🎯 What it does: A high-precision paint deposition model considering the spray gun's position and orientation was developed. The model's validity was verified by separately modeling the impact angle and spray distance, and by proposing a special function to map 3D vectors to a 2D coordinate system.

High stimuli virtual reality training for a brain controlled robotic wheelchair

Alexander Thomas, Tom Carlson

ClassificationRobotic IntelligenceBiomedical Data

🎯 What it does: A comparative study on training methods for a brain-computer interface (BCI)-controlled robotic wheelchair, where 15 participants were trained in simple tasks, a two-dimensional display virtual environment (VR-2DD), and a virtual environment using a VR headset (VR-HMD) with high and low noise mixed training, followed by online testing in the same virtual training environment.

High-Curvature, High-Force, Vine Robot for Inspection

Mijaíl Jaén Mendoza, E. Hawkes

Robotic Intelligence

🎯 What it does: Designed and implemented a vine robot capable of achieving high curvature and high force in confined environments, utilizing anisotropic extension of composite corrugated films for plant-inspired steering.

High-Degrees-of-Freedom Dynamic Neural Fields for Robot Self-Modeling and Motion Planning

Lennart Schulze, Hod Lipson

Robotic IntelligenceNeural Radiance FieldImage

🎯 What it does: Learn self-modeling for robot motion planning using neural field models from only 2D images, camera poses, and robot configuration data.

High-Dimensional Controller Tuning through Latent Representations

Alireza Sarmadi, F. Khorrami

OptimizationRepresentation LearningHyperparameter SearchReinforcement Learning

🎯 What it does: Propose a method that maps high-dimensional controller parameters to a low-dimensional latent space, and combines Bayesian optimization within the actor-critic framework to automatically and efficiently adjust controller parameters.

High-speed interfacial flight of an insect-scale robot

Hang Gao, E. Helbling

Robotic IntelligencePhysics Related

🎯 What it does: Designed and implemented a bug-scale robot γ-bot that generates thrust at the air-water interface using wings and supports its body via three passive legs utilizing surface tension, achieving high-speed flight and turning on water; simultaneously constructed and validated a simplified drag force model to estimate the robot's speed; experiments showed the robot can reach a maximum speed of 0.9 m/s, perform left and right turns, and carry an additional 419 mg load;

High-throughput Visual Nano-drone to Nano-drone Relative Localization using Onboard Fully Convolutional Networks

Luca Crupi, D. Palossi

Pose EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Developed a vision-based fully convolutional neural network (FCNN) using low-resolution grayscale cameras, achieving relative pose estimation between small nano-drones, with 39 Hz real-time inference implemented on a single ultra-low power system-on-chip (SoC);

Highway-Driving with Safe Velocity Bounds on Occluded Traffic

Truls Nyberg, Jana Tumova

Autonomous Driving

🎯 What it does: Proposed a high-speed driving method in obstructed traffic scenarios that ensures safety without being overly cautious.

HIO-SDF: Hierarchical Incremental Online Signed Distance Fields

V. Vasilopoulos, Volkan Isler

Representation Learning

🎯 What it does: Proposes a method called HIO-SDF, which represents the environment as a Signed Distance Field (SDF) and combines a coarse voxel grid with high-resolution local information through a hierarchical structure, trained using a neural network to achieve online incremental updates.

History-Aware Planning for Risk-free Autonomous Navigation on Unknown Uneven Terrain

Yinchuan Wang, Chaoqun Wang

Autonomous Driving

🎯 What it does: Proposed a hierarchical systematic pipeline to achieve safe and fast navigation without a map on unknown uneven terrain.

How Does Perception Affect Safety: New Metrics and Strategy

Xiaotong Zhang, Kamal Youcef-Toumi

Autonomous DrivingComputational EfficiencyTransformer

🎯 What it does: The proposed method quantifies the relationship between perception metrics and execution metrics, introduces two safety perception indicators, Critical Collision Probability (CCP) and Average Collision Probability (ACP), and develops an attention processing strategy to enhance perception efficiency and safety.

How Many Views Are Needed to Reconstruct an Unknown Object Using NeRF?

Sicong Pan, Maren Bennewitz

Convolutional Neural NetworkNeural Radiance FieldMesh

🎯 What it does: Proposes a non-iterative view planning pipeline based on Predicted Required Views (PRV), utilizing PRVNet to predict the number of required views for the target object and planning a global shortest path to complete NeRF reconstruction.

How to Prompt Your Robot: A PromptBook for Manipulation Skills with Code as Policies

Montse Gonzalez Arenas, Andy Zeng

Robotic IntelligenceReinforcement Learning from Human FeedbackLarge Language ModelPrompt EngineeringChain-of-Thought

🎯 What it does: Proposes PromptBook, a collection of multiple prompting paradigms for generating code enabling robots to perform new operational skills.

How to Train Your Neural Control Barrier Function: Learning Safety Filters for Complex Input-Constrained Systems

Oswin So, Chuchu Fan

OptimizationSafty and PrivacyReinforcement Learning

🎯 What it does: This paper studies methods for training neural control barrier functions (NCBF) in nonlinear systems with input constraints and high relative order, and proposes a new framework—Policy Neural Control Barrier Function (PNCBF)—that constructs control barrier functions by learning the value function of a baseline policy. Its effectiveness is validated on simulation and hardware platforms.

HPF-SLAM: An Efficient Visual SLAM System Leveraging Hybrid Point Features

Xin Su, Eckehard G. Steinbach

Simultaneous Localization and MappingImageVideo

🎯 What it does: Integrate manually designed and learnable point features into a single SLAM system, and design preprocessing, intra-class feature matching, and a Hybrid Bag-of-Words model;

HPL-ViT: A Unified Perception Framework for Heterogeneous Parallel LiDARs in V2V

Yuhang Liu, Fei Wang

Data SynthesisAutonomous DrivingTransformerPoint Cloud

🎯 What it does: Proposed a unified perception framework HPL-ViT, built a hardware prototype, and developed a heterogeneous LiDAR dataset OPV2V-HPL for LiDAR network information sharing in V2V environments.

HR-APR: APR-agnostic Framework with Uncertainty Estimation and Hierarchical Refinement for Camera Relocalisation

Changkun Liu, Tristan Braud

Pose EstimationComputational EfficiencySimultaneous Localization and MappingImage

🎯 What it does: Proposes a framework named HR-APR that is independent of specific camera pose regressors. It estimates uncertainty using the cosine similarity between query and database features and enhances pose accuracy through hierarchical refinement.

HSPNav: Hierarchical Scene Prior Learning for Visual Semantic Navigation Towards Real Settings

Jiaxu Kang, Jianxin Wang

Autonomous DrivingRobotic IntelligenceReinforcement LearningVision-Language-Action ModelImageGraphRetrieval-Augmented Generation

🎯 What it does: Propose a visual semantic navigation strategy HSPNav based on hierarchical scene priors and deep reinforcement learning;

Human Observation-Inspired Trajectory Prediction for Autonomous Driving in Mixed-Autonomy Traffic Environments

Haicheng Liao, Chengzhong Xu

Autonomous DrivingConvolutional Neural NetworkGraph Neural NetworkVideoTime Series

🎯 What it does: Proposed a trajectory prediction model based on human observation behavior for mixed autonomous driving traffic environments

Human Preference-aware Rebalancing and Charging for Shared Electric Micromobility Vehicles

Heng Tan, Yu Yang

OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTabularTime Series

🎯 What it does: Proposed and implemented PERCEIVE, a rebalancing and charging framework for shared electric micro-vehicles that considers human preferences.

Human Robot Shared Control in Surgery: A Performance Assessment

Longrui Chen, F. Baena

Robotic IntelligenceReinforcement Learning

🎯 What it does: Analyze various methods of human-robot shared control in surgery, and propose the tGAIL reinforcement learning algorithm and a shared control scheme with dynamic priority adjustment based on deviation distance

Human-Aligned Longitudinal Control for Occluded Pedestrian Crossing With Visual Attention

Vinal Asodia, Saber Fallah

Autonomous DrivingReinforcement LearningImage

🎯 What it does: Propose an adaptive reward function that utilizes visual attention maps to detect pedestrians in driving scenarios, dynamically switching the priority between safety or efficiency based on current observations, thereby achieving human-value-oriented longitudinal control in occluded pedestrian crossing scenarios.

Human-Centered Autonomy for UAS Target Search

Hunter M. Ray, nisar. ahmed

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a human-machine centered autonomous framework that infers geospatial task context through dynamic feature sets and guides a probabilistic goal search planner.

Human-Exoskeleton Locomotion Interaction Experience Transfer: Speeding up and Improving the Performance of Preference-based Optimizations of Exoskeleton Assistance During Walking

Hongwu Li, Yanhe Zhu

Optimization

🎯 What it does: Proposed a preference-based human-computer interaction experience transfer framework that leverages experienced users' experiences to accelerate and enhance novice users' exoskeleton-assisted parameter optimization.

Human-Robot Complementary Collaboration for Flexible and Precision Assembly

Shichen Cao, Jing Xiao

Robotic Intelligence

🎯 What it does: Propose a human-robot collaborative precision assembly method that achieves flexibility and efficiency in tight fitting assembly of complex-shaped components under uncertainty conditions, without requiring human assembly skills or robot knowledge, by leveraging the strengths of human natural abilities and robot systems.

Human-Robot Gym: Benchmarking Reinforcement Learning in Human-Robot Collaboration

Jakob Thumm, Matthias Althoff

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: Proposes the Human-Robot Gym benchmark suite for human-robot collaborative safe reinforcement learning, providing a modular simulation framework and six challenging tasks.

Human-Robot Interactive Creation of Artistic Portrait Drawings

Fei Gao, Peng Li

RestorationSegmentationDepth EstimationRobotic IntelligenceImage

🎯 What it does: Proposed a human-robot interactive art portrait drawing system (HRICA), enabling humans and robots to alternate in drawing strokes on a canvas, and achieving collaborative creation through the robot's understanding of human intent.

HumanMimic: Learning Natural Locomotion and Transitions for Humanoid Robot via Wasserstein Adversarial Imitation

Annan Tang, M. Inaba

Robotic IntelligenceReinforcement LearningGenerative Adversarial Network

🎯 What it does: Proposed a system based on Wasserstein adversarial imitation learning, enabling humanoid robots to mimic human movements, replicate natural full-body walking patterns, and achieve seamless transitions.

Hyblock: Hardware Realization and Control of Modular Hydraulic Robots with Dowel Connectors

S. Hyon, Yasushi Saitou

Robotic Intelligence

🎯 What it does: Designed and implemented the Hyblock modular hydraulic robot, which includes a C-type expandable wooden pivot docking mechanism and an MHSB hydraulic circuit. Experiments were conducted on pressure-based torque control and magnetic sensor docking control. Subsequently, a dynamic reconfiguration and work space motion control framework based on wooden pivot connectors was proposed. Simulation verification demonstrated the ability of collective modular robots to complete target motion tasks while maintaining normal contact force of the connectors.

Hybrid Force-Position Control of an Elastic Tendon-Driven Scrubbing Robot (TEDSR)

Noah Harmatz, Aaron D. Mazzeo

Robotic Intelligence

🎯 What it does: Implement hybrid force-displacement control using elastic rope-driven scrubbing robots to study the effect of joint stiffness on force control interference suppression during the scrubbing process.

Hybrid Robot for Percutaneous Needle Intervention Procedures: Mechanism Design and Experiment Verification

H. Zhang, Fuchun Sun

Robotic Intelligence

🎯 What it does: Designed and verified a 6-degree-of-freedom hybrid robot for percutaneous acupuncture intervention.

Hybrid Volitional Control of a Robotic Transtibial Prosthesis using a Phase Variable Impedance Controller

Ryan R. Posh, Patrick M. Wensing

Robotic IntelligenceBiomedical Data

🎯 What it does: The study investigates the use of a phase variable impedance controller (PVIC) defined by full knee kinematics and a hybrid voluntary control framework (PVI-HVC) on a robotic transtibial prosthesis, comparing three control strategies in terms of gait progress estimation and biomechanical performance on a healthy human subject.

Hydrodynamic Interactions in Schooling Fish: Prioritizing Real Fish Kinematics Over Travelling-wavy Undulation

Li-Ming Chao, Liang Li

Physics Related

🎯 What it does: The influence of real fish kinematics and idealized traveling wave motions on fluid dynamics interactions in fish schools was studied using computational fluid dynamics (CFD).

HyperLeg: Biomechanics-Inspired High-DOF Leg and Toe Mechanism for Highly Dynamic Motions

Do-Yun Kim, Yong-Jae Kim

Robotic Intelligence

🎯 What it does: Developed a leg mechanism inspired by human biomechanics, featuring a one-degree-of-freedom knee joint, two-degree-of-freedom ankle joint, and one-degree-of-freedom toe joint. All actuators are centralized near the proximal thigh to reduce distal mass; high-load belts and unique linkages achieve high reverse drive capability and joint stiffness. Actuator force coupling design enables support for high-dynamic motions such as jumping.

HyperPPO: A scalable method for finding small policies for robotic control

Shashank Hegde (University of Southern California), G. Sukhatme (University of Southern California)

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose the HyperPPO algorithm, which uses graph hypernetworks to estimate the weights of multiple neural network architectures in a single training process, thereby obtaining multiple small and high-performance control strategies, and implementing decentralized control on Crazyflie2.1 quadrotors.

i-Octree: A Fast, Lightweight, and Dynamic Octree for Proximity Search

Jun Zhu, T. Zhang

Computational EfficiencyPoint Cloud

🎯 What it does: Proposed a dynamic octree called i-Octree for fast nearest neighbor search and real-time dynamic point cloud updates.

IBBT: Informed Batch Belief Trees for Motion Planning Under Uncertainty

Dongliang Zheng, P. Tsiotras

Autonomous Driving

🎯 What it does: Proposes the IBt motion planning algorithm addressing motion and perception uncertainties

iBoW3D: Place Recognition Based on Incremental and General Bag of Words in 3D Scans

Yuxiaotong Lin, Liang Li

RecognitionSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a 3D point cloud localization method based on an incremental generic Bag-of-Words model, combining adaptable key points with 3D local feature extraction. The incremental BoW model is used to achieve coarse-to-fine candidate screening of the database, and geometric verification is applied to identify loops, while a supplementary metric is introduced to address the omissions of traditional metrics.

ICGNet: A Unified Approach for Instance-Centric Grasping

René Zurbrügg, Fisher Yu

Object DetectionRobotic IntelligencePoint Cloud

🎯 What it does: Proposes an end-to-end, instance-oriented grasping architecture that generates instance-centric representations for each locally visible object from single-view point clouds, enabling reconstruction and grasping detection in cluttered table-top scenes.

IDD-X: A Multi-View Dataset for Ego-relative Important Object Localization and Explanation in Dense and Unstructured Traffic

Chirag Parikh, R. Sarvadevabhatla

Autonomous DrivingExplainability and InterpretabilityVideo

🎯 What it does: Proposed a dual-view driving video dataset (IDD-X) for dense and unordered traffic scenarios, and designed a deep network model for multi-target localization and explanation prediction for each target.

IFFNeRF: Initialisation Free and Fast 6DoF pose estimation from a single image and a NeRF model

Matteo Bortolon, A. D. Bue

Pose EstimationNeural Radiance FieldImage

🎯 What it does: Utilize the NeRF model with Metropolis-Hasting sampling of surface points, combined with pixel-level view synthesis and attention mechanisms, to solve the 6DoF camera pose from a single image.

IKLink: End-Effector Trajectory Tracking with Minimal Reconfigurations

Yeping Wang, M. Gleicher

OptimizationRobotic IntelligenceGraph Neural Network

🎯 What it does: Propose an IKLink method to minimize the number of reconfigurations when a robot's end-effector performs trajectory tracking.

Implementation and Assessment of an Augmented Training Curriculum for Surgical Robotics

Alberto Rota, E. Momi

Robotic Intelligence

🎯 What it does: Developed and validated a VR simulator with haptic enhancement containing 8 surgical tasks;

Implicit Coarse-to-Fine 3D Perception for Category-level Object Pose Estimation from Monocular RGB Image

Jia Li, Xueying Qin

Pose EstimationImage

🎯 What it does: Proposes an end-to-end framework for category-level pose estimation from a single RGB image, named Feature Auxiliary Perception Network (FAP-Net).

Implicit Neural Representations for Breathing-compensated Volume Reconstruction in Robotic Ultrasound

Yordanka Velikova, N. Navab

RestorationRobotic IntelligenceNeural Radiance FieldBiomedical DataUltrasound

🎯 What it does: Introduces a 3D reconstruction method using implicit neural representations for respiratory compensation in robotic ultrasound systems, enhancing the automation and quality of abdominal ultrasound imaging;

Improved M4M: Faster and Richer Planning for Manipulation Among Movable Objects in Cluttered 3D Workspaces

D. Saxena, Maxim Likhachev

Robotic Intelligence

🎯 What it does: Improve E-M4M to propose I-M4M, achieving faster and more diverse movable object planning.

Improving Autonomous Driving Safety with POP: A Framework for Accurate Partially Observed Trajectory Predictions

Sheng Wang, Ming Liu

Autonomous DrivingKnowledge DistillationSequential

🎯 What it does: Proposed and implemented a framework called POP for trajectory prediction in congested urban road scenarios with partial observations.

Improving Neural Indoor Surface Reconstruction with Mask-Guided Adaptive Consistency Constraints

Xinyi Yu, Linlin Ou

GenerationImageMesh

🎯 What it does: This paper proposes a two-stage training process that separates perspective-related and irrelevant colors, and utilizes two novel consistency constraints and a masking scheme to enhance the quality of indoor scene reconstruction based on neural implicit surfaces.

Improving Offline Reinforcement Learning with Inaccurate Simulators

Yiwen Hou, Feng Wu

Data SynthesisReinforcement LearningGenerative Adversarial Network

🎯 What it does: Propose a new method that combines offline datasets with inaccurate simulator data. The method pre-trains a GAN to fit the state distribution of offline data, then samples simulator data starting from the generator's distribution, and reweights the sampled data using the discriminator to improve offline reinforcement learning performance.

Improving Out-of-Distribution Generalization of Learned Dynamics by Learning Pseudometrics and Constraint Manifolds

Yating Lin, Dmitry Berenson

Robotic IntelligenceContrastive LearningPhysics Related

🎯 What it does: Proposes a method to improve the prediction accuracy of robot dynamics models in out-of-distribution (OOD) states by learning pseudo-metrics and constraint manifolds.

Improving Radial Imbalances with Hybrid Voxelization and RadialMix for LiDAR 3D Semantic Segmentation

Jiale Li, Yong Ding

SegmentationPoint Cloud

🎯 What it does: Propose the Hi-VoxelNet network, utilizing hybrid voxelization and RadialMix data augmentation to address the radial imbalance issue in LiDAR 3D semantic segmentation.

Improving the Generalization of Unseen Crowd Behaviors for Reinforcement Learning based Local Motion Planners

Wen Zheng Terence Ng, Tianwei Zhang

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose a method that enhances agent diversity within a single strategy by maximizing an information-theoretic objective, to improve the generalization capability of RL local motion planners for unseen crowd behaviors.

Improving the ROS 2 Navigation Stack with Real-Time Local Costmap Updates for Agricultural Applications

Ettore Sani, Stefano Carpin

SegmentationRobotic IntelligenceImageAgriculture Related

🎯 What it does: A lightweight method was developed that uses a depth camera to perform pixel-level classification of images and injects corrections in real-time into Nav2's local cost map to improve robot navigation in agricultural environments.

IMU-Aided Event-based Stereo Visual Odometry

Junkai Niu, Yi Zhou

Pose EstimationGenerative Adversarial NetworkSimultaneous Localization and MappingOptical Flow

🎯 What it does: Improved the previous direct event-based stereo visual odometry pipeline to enhance accuracy and efficiency.

In-Hand Rolling Manipulation Based on Ball-on-Cloth System

Hinano Ichikura, Mitsuru Higashimori

Robotic Intelligence

🎯 What it does: Proposed a ball-cloth system based on fingertip cloth patches, utilizing the flexible deformation of the cloth to control the rolling of the ball, achieving in-hand rolling manipulation.

Incipient Slip-Based Rotation Measurement via Visuotactile Sensing During In-Hand Object Pivoting

Mingxuan Li, Yao Jiang

Robotic IntelligenceMultimodality

🎯 What it does: A generalized 2D contact model was proposed, and a rotation measurement method based on line features in the adhesion area was designed, applied to the Tac3D tactile sensor for measuring object rotation in hand.

Inconstant curvature kinematics of parallel continuum robot without static model

Tao Zhang, Hongliang Ren

Robotic Intelligence

🎯 What it does: This study proposes and verifies a more accurate kinematic model without requiring a static model, focusing on the motion advantages of mini parallel continuum robots after passing through narrow curved channels;

Incorporating Scene Graphs into Pre-trained Vision-Language Models for Multimodal Open-vocabulary Action Recognition

Chao Wei, Zhidong Deng

RecognitionTransformerVision Language ModelVision-Language-Action ModelVideoTextMultimodalityGraph

🎯 What it does: Proposed the Action-SGFA method for aligning action features based on scene graphs, and designed a new training paradigm to enhance the performance of pre-trained vision-language models in multi-modal open-vocabulary action recognition.

Increasing SLAM Pose Accuracy by Ground-to-Satellite Image Registration

Yanhao Zhang, Hongdong Li

Simultaneous Localization and MappingImage

🎯 What it does: A method fusing visual SLAM with deep learning-based ground-to-satellite image registration to enhance localization accuracy.

Increasing the Absolute Position Accuracy of Industrial Robots by Means of a Deep Continual Evidential Regression Model *

E. Uhlmann, Gang Wang

Robotic Intelligence

🎯 What it does: Propose a data-driven method considering kinematics, elasticity, and thermal effects, using continual learning algorithms to achieve process-parallel training, avoid catastrophic forgetting, and provide confidence intervals for predictions, thereby improving the absolute position accuracy of industrial robots.

Incremental 3D Reconstruction through a Hybrid Explicit-and-Implicit Representation

Feifei Li, Rui Huang

OptimizationRepresentation LearningNeural Radiance Field

🎯 What it does: Achieving incremental 3D reconstruction through the hybrid use of explicit and implicit representations

Incremental Learning of Full-Pose Via-Point Movement Primitives on Riemannian Manifolds

Tilman Daab, Tamim Asfour

Robotic IntelligenceSequential

🎯 What it does: This paper proposes seven basic operations for incremental learning, improvement, and reorganization of robot motion primitives (MP) library, and provides explicit formulas for five spatial operations in the library composed of via-point motion primitives (VMP); incremental learning of position and orientation VMP parameters is achieved through Riemannian manifold theory while maintaining a fixed number of parameters; finally, the method is evaluated using sequentially provided motion capture data.

Indoor Exploration and Simultaneous Trolley Collection Through Task-Oriented Environment Partitioning

Junjie Gao, M. Meng

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Proposes a framework that simultaneously performs exploration and pallet search, leveraging task-oriented environmental partitioning to construct a hybrid map, followed by generating feasible paths through TSP-PC and topological graph search to achieve indoor exploration and pallet collection.

Inexpensive, Automated Pruning Weight Estimation in Vineyards

Jonathan Jaramillo, Kirstin Petersen

ImageAgriculture Related

🎯 What it does: A low-cost and simple computer vision method is proposed to measure pruning weight in vineyards using smartphone cameras and structured light.

Infer and Adapt: Bipedal Locomotion Reward Learning from Demonstrations via Inverse Reinforcement Learning

Feiyang Wu, Ye Zhao

Robotic IntelligenceReinforcement Learning

🎯 What it does: Utilize inverse reinforcement learning (IRL) to learn the expert reward function, and train a bipedal walking strategy based on this reward to enhance walking performance on unknown complex terrains.

Influence of Camera-LiDAR Configuration on 3D Object Detection for Autonomous Driving

Ye Li, Ding Zhao

Object DetectionAutonomous DrivingMultimodalityPoint Cloud

🎯 What it does: Proposes an information theory-based evaluation metric for camera-LiDAR configuration, and constructs an accelerated framework on the CARLA simulator for data collection, model training, and performance evaluation, exploring the impact of sensor configurations on the performance of learning-based 3D object detection.

Information-driven Affordance Discovery for Efficient Robotic Manipulation

Pietro Mazzaglia, D. Dijkman

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose an information-based affordance learning method (IDA) for operations such as grasping, stacking, and pulling out drawers, accelerating the discovery of visual affordances through guided interaction, and validating it in simulation and real robots.

Informed Reinforcement Learning for Situation-Aware Traffic Rule Exceptions

Daniel Bogdoll, J. M. Zöllner

Autonomous DrivingReinforcement Learning

🎯 What it does: Introduced information-based reinforcement learning, using structured rule manuals as knowledge sources, and designed a context-aware reward mechanism to enable learning of traffic rule anomalies in complex autonomous driving scenarios.

Instructing Robots by Sketching: Learning from Demonstration via Probabilistic Diagrammatic Teaching

Weiming Zhi, M. Johnson-Roberson

Robotic IntelligenceImageTime Series

🎯 What it does: Proposes a Diagrammatic Teaching paradigm, allowing users to draw示范 trajectories on 2D scene graphs, enabling robots to learn motion trajectories in 3D task spaces through generative models; implements the Ray-Tracing Probabilistic Trajectory Learning (RPTL) framework to extract time-varying probability density, perform ray-tracing mapping, fit probabilistic models of motion trajectories, and ultimately generate new imitation trajectories; validates the approach in simulations and on real robots (fixed-base robotic arms and arms mounted on quadruped robots).

Integrated Data-driven Inference and Planning-based Human Motion Prediction for Safe Human-Robot Interaction

Youngim Nam, Cheolhyeon Kwon

Autonomous DrivingOptimizationSafty and PrivacyRecurrent Neural Network

🎯 What it does: Propose a unified prediction and planning algorithm that enables autonomous vehicles to safely interact with uncertain human-driven vehicles.

Integrating Open-World Shared Control in Immersive Avatars

Patrick Naughton, Kris Hauser

Robotic IntelligenceWorld Model

🎯 What it does: Proposed a framework that integrates open-world shared control into immersive avatar robots

Integrating Predictive Motion Uncertainties with Distributionally Robust Risk-Aware Control for Safe Robot Navigation in Crowds

Kanghyun Ryu, Negar Mehr

Autonomous DrivingRobotic Intelligence

🎯 What it does: Proposes a distributionally robust probabilistic constrained model predictive control (DRCC-MPC) framework to achieve safe robot navigation among crowds.

Intelligent Disinfection Robot with High-Touch Surface Detection and Dynamic Pedestrian Avoidance

Yunfei Luan, Yao Guo

Robotic IntelligenceImagePoint Cloud

🎯 What it does: Designed a disinfection robot system with high-touch surface detection and dynamic pedestrian avoidance capabilities, equipped with a mobile platform, RGB-D camera, and spraying device;

Intelligent Mode-switching Framework for Teleoperation

Burak Kizilkaya, Muhammad Ali Imran

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes an intelligent mode-switching framework that combines user intent recognition and deep reinforcement learning (DRL) to achieve seamless switching between teleoperation and autonomous modes.

InteRACT: Transformer Models for Human Intent Prediction Conditioned on Robot Actions

K. Kedia, Sanjiban Choudhury

Robotic IntelligenceTransformerSupervised Fine-Tuning

🎯 What it does: Propose the InteRACT model, which leverages a Transformer pre-trained on large-scale human-human interaction data, followed by fine-tuning on a limited amount of human-robot interaction data to achieve conditional human intent prediction, and introduces teleoperation technology using a 7-DoF robotic arm to collect diverse human-robot collaborative manipulation data.