ICRA 2024 Papers — Page 7
IEEE International Conference on Robotics and Automation · 1760 papers
Fast and Consistent Covariance Recovery for Sliding-window Optimization-based VINS
Chuchu Chen, Guoquan Huang
Autonomous DrivingOptimizationSimultaneous Localization and Mapping
🎯 What it does: Proposes a technique to achieve consistent and fast covariance recovery in a sliding window optimization-based visual inertial navigation system.
Fast and Robust Normal Estimation for Sparse LiDAR Scans
Igor Bogoslavskyi, Raymond Phan
Autonomous DrivingComputational EfficiencyPoint Cloud
🎯 What it does: Propose a normal estimation method that utilizes the known emission pattern of a mechanical LiDAR scanner, labels points and their neighbors based on the angle of connecting lines, and estimates surface normals using only neighbors with the same label to avoid over-smoothing in high-curvature regions.
Fast and Robust Point Cloud Registration with Tree-based Transformer
Guangyan Chen (Beijing Institute of Technology), Yufeng Yue (Beijing Institute of Technology)
Pose EstimationTransformerPoint Cloud
🎯 What it does: Proposed Tree-based Transformer (TrT) for point cloud registration, constructing a coarse-to-fine feature tree and using Tree-based Attention (TrA) to progressively focus on key points, achieving rich local and global feature extraction with linear complexity.
Fast Photoacoustic Microscopy with Robot Controlled Microtrajectory Optimization
Yating Luo, Guang-Zhong Yang
OptimizationRobotic IntelligenceBiomedical Data
🎯 What it does: Propose a robotic control-based micro-trajectory optimized fast photoacoustic microscopy scheme
Fast Wheeled Driving to Legged Leaping onto a Step in a Leg-Wheel Transformable Robot
Zhi-Ren Chen, Pei-Chun Lin
OptimizationRobotic Intelligence
🎯 What it does: Studies the dynamic control strategy of leg-wheel morphing robots rapidly switching from wheeled driving mode to legged jumping mode to leap over steps higher than themselves and then smoothly return to wheeled mode.
FastOcc: Accelerating 3D Occupancy Prediction by Fusing the 2D Bird’s-Eye View and Perspective View
Jiawei Hou, Jian Pu
Autonomous DrivingComputational EfficiencyConvolutional Neural NetworkImageBenchmark
🎯 What it does: Propose the FastOcc method, which integrates 2D BEV and perspective views to enhance the speed of 3D occupancy prediction.
Fault Tolerant Neural Control Barrier Functions for Robotic Systems under Sensor Faults and Attacks
Hongchao Zhang, R. Poovendran
Robotic Intelligence
🎯 What it does: Proposed and implemented a fault-tolerant neural control barrier function (FT-NCBF) applicable to environments with sensor faults and attacks, learned the function using a data-driven approach, and subsequently designed control inputs while formally proving its safety.
FBPT: A Fully Binary Point Transformer
Zhixing Hou, Yan Yan
ClassificationRecognitionTransformerPoint Cloud
🎯 What it does: Designed a fully binarized point cloud Transformer model called FBPT, compressing weights and activations to significantly reduce model size and computational demands.
FC-Planner: A Skeleton-guided Planning Framework for Fast Aerial Coverage of Complex 3D Scenes
Chen Feng, Boyu Zhou
OptimizationRobotic IntelligencePoint CloudMesh
🎯 What it does: Proposes FC-Planner, a skeleton-guided planning framework that can quickly perform UAV coverage path planning in complex 3D scenarios without preprocessing.
FE-DeTr: Keypoint Detection and Tracking in Low-quality Image Frames with Events
Xiangyuan Wang, Huai Yu
Object DetectionObject TrackingPose EstimationTransformerImageMultimodality
🎯 What it does: A keypoint detection and tracking method called FE-DeTr, which integrates image frames and event streams, was developed. It achieves stable and efficient keypoint detection through time response consistency supervision and employs a spatiotemporal nearest neighbor search strategy for robust tracking.
Few-Shot Fruit Segmentation via Transfer Learning
Jordan A. James, William J. Beksi
SegmentationMeta LearningBenchmarkAgriculture Related
🎯 What it does: Developed a few-shot fruit segmentation framework based on transfer learning for semantic segmentation of field fruits.
Few-Shot Learning of Force-Based Motions From Demonstration Through Pre-training of Haptic Representation
Marina Y. Aoyama, S. Vijayakumar
Representation LearningRobotic IntelligenceSupervised Fine-TuningAuto Encoder
🎯 What it does: Propose a semi-supervised learning demonstration method, splitting the model into a pre-trained tactile representation encoder and a motion generation decoder trained with a small number of demonstrations, applied to sponge wiping tasks involving objects with different hardness and surface friction.
Few-Shot Panoptic Segmentation With Foundation Models
Markus Kappeler, Abhinav Valada
SegmentationRepresentation LearningTransformerImage
🎯 What it does: Proposes a lightweight network head-based few-shot annotated panoptic segmentation method SPINO, leveraging task-agnostic image features to generate high-quality pseudo labels under the condition of only ten annotated images;
FF-LOGO: Cross-Modality Point Cloud Registration with Feature Filtering and Local to Global Optimization
Nan Ma, Y. Liu
OptimizationPoint Cloud
🎯 What it does: Proposed a cross-modal point cloud registration framework named FF-LOGO, combining feature filtering with local-global optimization;
Field Evaluation of a Prioritized Path-Planning Algorithm for Heterogeneous Agricultural Tasks of Multi-UGVs
Yuseung Jo, H. Son
Simultaneous Localization and MappingAgriculture Related
🎯 What it does: A priority-based path planning algorithm is proposed for multiple heterogeneous unmanned ground vehicles (UGV) in agricultural operations (harvesting and transportation).
Field-evaluated Closed Structure Soft Gripper Enhances the Shelf Life of Harvested Blackberries
Philip H. Johnson, Marcello Calisti
Robotic IntelligenceAgriculture Related
🎯 What it does: Developed the first closed-structure soft robotic gripper for blackberry harvesting, initially tested on sensorized tomato physical twins, then tested in a farm's multi-layer covered structure following grower-guided protocols, and compared with professional human pickers.
Field-VIO: Stereo Visual-Inertial Odometry Based on Quantitative Windows in Agricultural Open Fields
Jianjing Sun, Junming He
Simultaneous Localization and MappingImageAgriculture Related
🎯 What it does: Proposes a stereo visual-inertial odometry system based on ORB-SLAM3, describing the robot's crop-following trajectory along quantitative window concepts and driving state quantization algorithms, utilizing parallel constraints to construct spatial constraints, and performing global pose correction on abnormal window keyframes to reduce cumulative errors.
FIMP: Future Interaction Modeling for Multi-Agent Motion Prediction
Sungmin Woo, Sangyoun Lee
Autonomous DrivingSequential
🎯 What it does: Proposed the FIMP framework, which implicitly extracts potential future information from intermediate feature layers using a future decoder, and identifies interacting entity pairs through future affinity learning and top-k filtering strategies to achieve multi-agent motion prediction.
Fine-Grained Pillar Feature Encoding Via Spatio-Temporal Virtual Grid for 3D Object Detection
Konyul Park, J. Choi
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: Developed a Fine-Grained Pillar Feature Encoding (FG-PFE) architecture that utilizes a Spatio-Temporal Virtual (STV) grid to encode LiDAR point clouds in vertical, temporal, and horizontal dimensions with fine-grained distribution, and enhances 3D object detection performance through attention aggregation.
Fine-Tuning Point Cloud Transformers with Dynamic Aggregation
Jiajun Fei, Zhidong Deng
SegmentationComputational EfficiencyTransformerAuto EncoderPoint Cloud
🎯 What it does: Propose a dynamic aggregation method to replace static aggregation (e.g., average or max pooling) during the fine-tuning of pre-trained point cloud Transformers
Fingertip Ultrasonic Array for Tactile Rendering
Jace Rozsa, G. Fedder
Ultrasound
🎯 What it does: Developed a miniature tactile stimulation device using focused ultrasound to generate tactile sensations on fingers;
Fit-NGP: Fitting Object Models to Neural Graphics Primitives
Marwan Taher, Andrew J. Davison
Pose EstimationRobotic IntelligenceNeural Radiance Field
🎯 What it does: Demonstrates that density fields generated by state-of-the-art efficient radiance field reconstruction methods can be used for high-precision and robust pose estimation of known 3D model objects, and proposes a fully automated robotic arm system using only a single wrist-mounted camera, capable of scanning scenes, detecting, and estimating the 6-degree-of-freedom (6-DOF) poses of multiple objects within minutes.
Fitting Parameters of Linear Dynamical Systems to Regularize Forcing Terms in Dynamical Movement Primitives
F. Stulp, Carme Torras
Optimization
🎯 What it does: Propose a new dynamic movement primitive (DMP) formulation that uses a generalized logistic function as the delayed target system, and regularizes the forcing term and improves interpolation accuracy by automatically fitting the parameters of the linear dynamical system through optimization.
Fixture calibration with guaranteed bounds from a few correspondence-free surface points
R. Haugaard, Thorbjørn Mosekjær Iversen
Pose EstimationPoint Cloud
🎯 What it does: Proposes a correspondence-free registration method where users only need to measure a few surface points, and the method provides a tight upper bound set of poses that satisfy these points.
Flexible Omnidirectional Driving Gear Mechanism with Adaptation over Arbitrary Curvatures
Moses Gladson Selvamuthu, R. Tadakuma
🎯 What it does: Developed a dual-degree-of-freedom omni-directional driving gear support structure for flexible displays (e.g., OLED or flexible LED);
Flight Validation of a Global Singularity-Free Aerodynamic Model for Flight Control of Tail Sitters
K. Murali, Leandro R. Lustosa
Optimization
🎯 What it does: A singularity-free aerodynamic model (ϕ-theory) for dual-engine tail-sitting flying wing vehicles was validated through flight tests, and its application in optimal control was demonstrated.
Floating-base manipulation on zero-perturbation manifolds
Brian Bittner, Kevin C. Wolfe
Robotic Intelligence
🎯 What it does: A motion planning method for floating base robots is proposed, achieving zero disturbance to the base by planning arm movements, and converting it into a nonholonomic RRT solution.
Flock-Formation Control of Multi-Agent Systems using Imperfect Relative Distance Measurements
Andreas Brandstätter, R. Grosu
Optimization
🎯 What it does: A distributed distance-based control (DDC) method was developed to control multi-agent systems in resource-constrained environments to achieve desired formations while simultaneously reaching target positions.
Flow Shadowing: A Method to Detect Multiple Flow Headings using an Array of Densely Packed Whisker-inspired Sensors
T. A. Kent, Sarah Bergbreiter
Robotic IntelligencePhysics Related
🎯 What it does: Direction detection of two simultaneous flow sources from different directions was achieved through densely arranged whisker-inspired flow sensors.
FLTRNN: Faithful Long-Horizon Task Planning for Robotics with Large Language Models
Jiatao Zhang, J. Gu
Robotic IntelligenceRecurrent Neural NetworkLarge Language ModelText
🎯 What it does: Propose the FLTRNN framework, which integrates task decomposition and memory management into LLM planning reasoning through a language-based RNN structure to enhance the fidelity and reliability of long-sequence task planning.
FocoTrack: Multi Object Tracking by Focusing On Overlap at Low Frame Rate
J. Lee, D. Chang
Object TrackingVideo
🎯 What it does: Proposed a multi-object tracking algorithm called FocoTrack, specifically designed for tracking at low frame rates, particularly effective in handling target overlaps.
FogROS2-Config: A Toolkit for Choosing Server Configurations for Cloud Robotics
∗. KaiyuanChen, Ken Goldberg
OptimizationBenchmark
🎯 What it does: Proposed the FogROS2-Config toolkit, which can automatically benchmark ROS2 nodes, quickly provide cloud computing service configurations that balance latency and cost, and validated its effectiveness in three categories of robotic applications: visual SLAM, grasping planning, and motion planning.
FogROS2-LS: A Location-Independent Fog Robotics Framework for Latency Sensitive ROS2 Applications
Kai-Peng Chen, Ken Goldberg
Robotic Intelligence
🎯 What it does: Propose the FogROS2-LS framework to achieve secure, location-agnostic connections with cloud/edge computing nodes, offloading traditional robot state estimation and feedback controllers to cloud and edge hardware without modifying existing ROS2 applications; dynamically identify and switch to the optimal service node meeting latency requirements in environments with multiple identical service nodes;
Follow the Footprints: Self-supervised Traversability Estimation for Off-road Vehicle Navigation based on Geometric and Visual Cues
Yurim Jeon, Seung-Woo Seo
Autonomous DrivingRobotic IntelligenceImage
🎯 What it does: Propose a self-supervised off-road feasibility estimation method based on geometric and visual information, using a Guided Filtering Network (GFN) and Footprint Supervision Module (FSM) to predict the robot's traversable area.
Foot Shape-Dependent Resistive Force Model for Bipedal Walkers on Granular Terrains
Xunjie Chen, Tao Liu
Robotic IntelligencePhysics Related
🎯 What it does: Proposed an improved resistance mechanics model and validated it through a large number of foot invasion experiments
Force Estimation at the Bionic Soft Arm’s Tool-center-point during the Interaction with the Environment
Samuel Pilch, Oliver Sawodny
Robotic Intelligence
🎯 What it does: Proposed and implemented a probabilistic force estimation model for the tool center point of a soft continuous robotic arm, and realized a hybrid force-displacement control based on this model;
Force Feedback Model-Predictive Control via Online Estimation
Armand Jordana, Ludovic Righetti
Optimization
🎯 What it does: Proposes an online estimation method that integrates force sensor feedback directly into nonlinear model predictive control (NMPC) to achieve precise force tracking.
Force-based semantic representation and estimation of feature points for robotic cable manipulation with environmental contacts
Andrea Monguzzi, A. Zanchettin
Representation LearningRobotic IntelligenceGraph Neural Network
🎯 What it does: A dual-arm robot equipped with dual-wrist torque sensors manipulates deformable linear objects (DLOs) in unknown environments. It proposes a strategy to estimate the pose of unknown environmental contact points, classify constraints (unidirectional, bidirectional, and fully constrained), and integrate semantic constraints into a graph-based DLO model. The model maintains accuracy when the DLO is under tension and dynamically updates during manipulation. Its feasibility is verified through simulation and real-world experiments.
ForceSight: Text-Guided Mobile Manipulation with Visual-Force Goals
Jeremy Collins, Charles C. Kemp
Robotic IntelligenceTransformerVision-Language-Action ModelImageText
🎯 What it does: Developed a text-guided mobile operating system called ForceSight, which utilizes a text-conditioned visual Transformer to predict visual-force targets, generating end-effector poses and force targets from a single RGBD image and text prompts to achieve mobile manipulation tasks.
Forgetting in Robotic Episodic Long-Term Memory
Joana Plewnia, Tamim Asfour
Robotic IntelligenceImageTime Series
🎯 What it does: Implemented and evaluated forgetting techniques in a robot's long-term memory, enabling merging of new information and optimizing memory content through fast filtering and slow precise offline pruning mechanisms
Frame Fusion with Vehicle Motion Prediction for 3D Object Detection
Xirui Li, Chaoxiang Ma
Object DetectionAutonomous DrivingImagePoint Cloud
🎯 What it does: Enhance 3D detection performance by projecting historical detection boxes 'forward' to the current frame and utilizing weighted non-maximum suppression fusion
From Bird’s-Eye to Street View: Crafting Diverse and Condition-Aligned Images with Latent Diffusion Model
Xiaojie Xu, Yingcong Chen
GenerationSupervised Fine-TuningDiffusion modelImage
🎯 What it does: Proposes a multi-view street scene image generation framework based on Bird’s-Eye View (BEV), which first converts BEV maps into corresponding multi-view semantic segmentation maps, and then uses these maps as conditions to guide a fine-tuned latent diffusion model in generating street scene images.
From Cooking Recipes to Robot Task Trees – Improving Planning Correctness and Task Efficiency by Leveraging LLMs with a Knowledge Network
M. Sakib, Yu Sun
OptimizationRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: Propose a pipeline that utilizes LLM to generate a cooking task tree, achieving correct and efficient robot cooking planning.
From Satellite to Ground: Satellite Assisted Visual Localization with Cross-view Semantic Matching
Xiyue Guo, Guofeng Zhang
Pose EstimationSimultaneous Localization and MappingImage
🎯 What it does: Propose a semantic-based cross-view localization method and integrate it into a SLAM system, utilizing satellite images for ground visual localization.
From Unstable Electrode Contacts to Reliable Control: A Deep Learning Approach for HD-sEMG in Neurorobotics
Eion Tyacke, S. F. Atashzar
Robotic IntelligenceConvolutional Neural NetworkBiomedical Data
🎯 What it does: Propose a deep learning model to ensure reliability for high-density surface electromyography (HD-sEMG) modules, applicable to neural robotic wearable interfaces.
FSD: Fast Self-Supervised Single RGB-D to Categorical 3D Objects
Mayank Lunayach, Muhammad Zubair Irshad
RecognitionPose EstimationImage
🎯 What it does: Propose a multi-stage training process that uses a single RGB-D image to achieve unlabeled 3D object recognition, predicting 3D shape, size, and 6D pose.
Fully 3D printable Robot Hand and Soft Tactile Sensor based on Air-pressure and Capacitive Proximity Sensing
Sean Taylor, Joohyung Kim
Robotic Intelligence
🎯 What it does: Designed and manufactured a low-cost, easy-to-build soft tactile sensing robotic hand that can be digitally fabricated, achieving force sensing and capacitive near-field sensing using 3D printing and off-the-shelf pressure and capacitive sensors.
Fully Distributed Shape Sensing of a Flexible Surgical Needle Using Optical Frequency Domain Reflectometry for Prostate Interventions
Jacynthe Francoeur, Samuel Kadoury
Robotic IntelligenceBiomedical DataComputed Tomography
🎯 What it does: Proposed a flexible surgical needle shape sensing method based on fully distributed grating sensors, using optical frequency domain reflectometry (OFDR) to model and measure the tip bending without prior knowledge of physical properties.
Fully Onboard Low-Power Localization with Semantic Sensor Fusion on a Nano-UAV using Floor Plans
Nicky Zimmerman, Luca Benini
Object DetectionComputational EfficiencyRobotic IntelligenceConvolutional Neural NetworkSimultaneous Localization and MappingImage
🎯 What it does: Propose an online, onboard nano UAV localization method that utilizes semantically annotated floor plans for localization;
FuncGrasp: Learning Object-Centric Neural Grasp Functions from Single Annotated Example Object
Hanzhi Chen, Stefan Leutenegger
Robotic IntelligenceImagePoint Cloud
🎯 What it does: Infer dense and reliable grasping configurations of unknown objects through classification priors using a single annotated object and single-view RGB-D observations, achieving the transfer of infinite grasping configurations via an object-centered continuous grasping function.
GAM-Depth: Self-Supervised Indoor Depth Estimation Leveraging a Gradient-Aware Mask and Semantic Constraints
Anqi Cheng, Kezhi Mao
Depth EstimationImage
🎯 What it does: Proposed the GAM-Depth model, which improves indoor self-supervised depth estimation using gradient-aware masks and semantic constraints to address depth inconsistency in textureless regions and depth differences at object boundaries.
GAMMA: Generalizable Articulation Modeling and Manipulation for Articulated Objects
Qiaojun Yu, Cewu Lu
Robotic IntelligenceMesh
🎯 What it does: Propose the GAMMA framework to learn joint modeling for multi-class robotic arm objects and adapt grasp poses, while improving manipulation performance by reducing modeling errors through adaptive operations.
GAMMA: Graspability-Aware Mobile MAnipulation Policy Learning based on Online Grasping Pose Fusion
Jiazhao Zhang, He Wang
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a mobile manipulation method based on online grasp pose fusion for grasp feasibility perception
Gas-Source Efficiently Active Searching in Unfamiliar Environments
Yue Zhai, Yanzi Miao
Robotic IntelligenceReinforcement LearningPoint Cloud
🎯 What it does: Proposed an end-to-end learning framework that integrates gas source localization with robot navigation, directly generating navigation actions using raw 3D-LiDAR observations and gas distribution information to achieve active gas source search;
Gathering Data from Risky Situations with Pareto-Optimal Trajectories
Brennan Brodt, Alyssa Pierson
Autonomous DrivingOptimization
🎯 What it does: Proposed a risk-aware path planning framework using multi-objective optimization, and developed methods to represent environmental information and risk in continuous monitoring scenarios. Trajectories that are Pareto-dominated in a recursive view are generated through local sampling.
Gaussian Mixture Likelihood-based Adaptive MPC for Interactive Mobile Manipulators
Dimitrios Rakovitis, Dennis Mronga
OptimizationRobotic Intelligence
🎯 What it does: A hierarchical adaptive MPC method based on Gaussian Mixture Models and Gaussian Mixture Regression is studied for predicting dynamic model parameters and accomplishing mobile manipulation tasks under multiple unknown environmental parameters.
Gaussian Process-based Traversability Analysis for Terrain Mapless Navigation
Abe Leininger, Lantao Liu
Autonomous DrivingOptimization
🎯 What it does: A geometry-based mapless navigation framework for rough terrain is proposed, combining sparse Gaussian process (SGP) local maps with rapidly exploring random tree* (RRT*) planner to generate high-resolution environmental representations and construct traversability maps for safe path planning.
Gaussian Process-Enhanced, External and Internal Convertible Form-Based Control of Underactuated Balance Robots
Feng Han, Jingang Yi
Robotic Intelligence
🎯 What it does: This study investigates and improves the externally and internally convertible (EIC) model-based control method for underactuated balancing robots. It first identifies potential loss-of-control conditions under EIC control, then proposes an enhanced EIC control scheme leveraging Gaussian Process (GP) data-driven robot dynamics models, and verifies its stability and performance through experiments.
Gaze-based Human-Robot Interaction System for Infrastructure Inspections
Sunwoong Choi, C. Yeum
Robotic Intelligence
🎯 What it does: Proposed a gaze-based mixed reality human-computer interaction system to enhance visual detection performance in infrastructure inspections;
GBEC: Geometry-Based Hand-Eye Calibration
Yihao Liu, Mehran Armand
Pose EstimationRobotic Intelligence
🎯 What it does: Proposes a geometry-based hand-eye calibration method (GBEC) to enhance the repeatability and accuracy of the transformation between the robot's end-effector and the sensor.
GelLink: A Compact Multi-phalanx Finger with Vision-based Tactile Sensing and Proprioception
Yuxiang Ma, Edward H. Adelson
Robotic IntelligenceImage
🎯 What it does: Proposes GelLink, a compact non-fully-driven articulated joint-driven robotic finger equipped with low-cost high-resolution visual tactile and proprioceptive sensors.
GelRoller: A Rolling Vision-based Tactile Sensor for Large Surface Reconstruction Using Self-Supervised Photometric Stereo Method
Zhiyuan Zhang, Hua Yang
Depth EstimationImagePoint Cloud
🎯 What it does: Designed a rolling cylindrical visual tactile sensor that achieves continuous and rapid perception over large surface areas through rolling, and proposed a self-supervised photometric stereo deep learning method. This method can obtain surface normals from a single frame image without prior calibration or stable illumination, followed by large-area surface reconstruction using normals and point cloud registration.
Gen2Sim: Scaling up Robot Learning in Simulation with Generative Models
Pushkal Katara, Katerina Fragkiadaki
Data SynthesisRobotic IntelligenceTransformerLarge Language ModelDiffusion modelChain-of-Thought
🎯 What it does: Automate the generation of 3D assets, task descriptions, task decomposition, and reward functions through large-scale generative models, enabling the scaling up of robot learning.
GenDOM: Generalizable One-shot Deformable Object Manipulation with Parameter-Aware Policy
So Kuroki, Yusuke Iwasawa
Robotic IntelligenceReinforcement LearningPoint CloudMesh
🎯 What it does: Propose a generic framework called GenDOM that can manipulate various deformable objects with only a single real demonstration.
Generalizable Thermal-based Depth Estimation via Pre-trained Visual Foundation Model
Ruoyu Fan, Wenping Wang
Depth EstimationTransformerImage
🎯 What it does: Propose a self-supervised method that leverages pre-trained RGB visual models to enhance thermal imaging depth estimation
Generalize by Touching: Tactile Ensemble Skill Transfer for Robotic Furniture Assembly
Hao-ming Lin, Ding Zhao
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes the Tactile Ensemble Skill Transfer (TEST) framework, which employs offline reinforcement learning with tactile feedback to address long-term and operational non-generalization issues in furniture assembly.
Generalized Correspondence Matching via Flexible Hierarchical Refinement and Patch Descriptor Distillation
Yu Han, Rui Fan
RetrievalKnowledge DistillationImage
🎯 What it does: To address the limitations of deep feature matching (DFM), a more flexible hierarchical refinement method is proposed, integrating patch descriptors and distillation strategies to enhance correspondence matching performance.
Generalized Partially Destructive Disassembly Planning for Robotic Disassembly
Malte Hansjosten, J. Fleischer
Robotic Intelligence
🎯 What it does: Proposes a general destructive disassembly planning method that can automatically derive destructive disassembly actions from symbolically represented disassembly states.
Generalizing Cooperative Eco-driving via Multi-residual Task Learning
V. Jayawardana, Kentaro Oguchi
Autonomous DrivingReinforcement Learning
🎯 What it does: Propose a Redundant Residual Task Learning (MRTL) framework that decomposes control tasks into nominal components addressed by traditional methods and residual terms addressed by deep reinforcement learning to achieve multi-scenario generalization.
Generalizing Trajectory Retiming to Quadratic Objective Functions
Gerry Chen, Seth Hutchinson
OptimizationRobotic Intelligence
🎯 What it does: Propose a trajectory re-timing algorithm based on factor graph variable elimination that can solve for the global optimal solution of a quadratic objective function.
Generating Environment-based Explanations of Motion Planner Failure: Evolutionary and Joint-Optimization Algorithms
Qishuai Liu, Martim Brandão
Autonomous DrivingOptimizationExplainability and Interpretability
🎯 What it does: Propose an automatically generated environment-based explanation method for motion planning failures
Generating robotic elliptical excisions with human-like tool-tissue interactions
Artūras Straižys, S. Ramamoorthy
Robotic Intelligence
🎯 What it does: The study utilizes a robot to learn elliptical resection techniques based on a few demonstrations, and enhances resection behavior through parameterized skill models and Bayesian optimization to align with expert scores and cutting force characteristics.
Generating Sparse Probabilistic Graphs for Efficient Planning in Uncertain Environments
Yasmin Veys, Nicholas Roy
OptimizationGraph
🎯 What it does: Generate a sparse probabilistic graph (road network) to support efficient planning in uncertain environments.
Generation of Steady Wheel Gait for Planar X-shaped Walker with Reaction Wheel
Fumihiko Asano, Cong Yan
Robotic IntelligencePhysics RelatedOrdinary Differential Equation
🎯 What it does: Studied and designed the Wheel Gait motion of a planar 3-degree-of-freedom X-shaped biped robot (equipped with reaction wheels), achieving stability of zero dynamics through control strategies, and conducted numerical generation and analysis of gaits for both linear and nonlinear models.
Generative Modeling of Residuals for Real-Time Risk-Sensitive Safety with Discrete-Time Control Barrier Functions
Ryan K. Cosner, A. Ames
OptimizationAuto Encoder
🎯 What it does: Train a state-conditioned generative model to represent the distribution of expected dynamics and actual system error residuals, and design a real-time risk-sensitive safety controller ORIO based on this.
Geometric Fabrics: a Safe Guiding Medium for Policy Learning
Karl Van Wyk, Nathan D. Ratliff
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed and implemented a safety-guided intermediate layer based on geometric fabrics, enabling reinforcement learning strategies to learn safer, nonlinear actions in a new behavioral dynamics space, and demonstrated the cube in-hand reorientation task on highly actuated robotic hands.
Geometric Slosh-Free Tracking for Robotic Manipulators
Jon Arrizabalaga, Markus Ryll
Robotic Intelligence
🎯 What it does: Propose a real-time oscillation-free tracking technique that generates an oscillation-free reference direction for the end-effector using a virtual quadrotor model, and implements feasible joint commands for a liquid transportation robot through PD control and RAC-based quadratic programming (QP).
Geometry-Informed Distance Candidate Selection for Adaptive Lightweight Omnidirectional Stereo Vision with Fisheye Images
Conner Pulling, S. Scherer
Depth EstimationAutonomous DrivingComputational EfficiencyImage
🎯 What it does: Propose a distance candidate selection method based on geometric information for adaptive lightweight panoramic stereo vision, aiming to reduce the number of hypothesis distance candidates in the cost volume.
GG-LLM: Geometrically Grounding Large Language Models for Zero-shot Human Activity Forecasting in Human-Aware Task Planning
M. Graule, Volkan Isler
Robotic IntelligenceTransformerLarge Language ModelMultimodality
🎯 What it does: Leveraging large language models (LLMs) to infer human next actions from multimodal information without fine-tuning, associating predicted actions with specific locations on a semantic map, and subsequently integrating these local activity predictions into a human perception task planner to significantly reduce adverse human-robot interaction incidents.
GIRA: Gaussian Mixture Models for Inference and Robot Autonomy
K. Goel, Wennie Tabib
Computational EfficiencyRobotic Intelligence
🎯 What it does: Proposed the open-source framework GIRA, which utilizes Gaussian Mixture Models (GMM) to achieve reconstruction, pose estimation, and occupancy modeling, while supporting low-bandwidth communication and high-resolution reconstruction;
Global Terminal Sliding Mode Control of Tethered Satellites Formation with Chattering Reduction via PID Laws
Bowen Su, Panfeng Huang
OptimizationPhysics Related
🎯 What it does: A global terminal sliding mode control method for tethered satellite systems under external disturbances is studied, and chattering in sliding mode control is reduced through PI/PD compensation.
Globalizing Local Features: Image Retrieval Using Shared Local Features with Pose Estimation for Faster Visual Localization
Wenzheng Song, Takayuki Okatani
Pose EstimationRetrievalImage
🎯 What it does: Propose a visual localization method that combines shared local features with pose estimation, using a single network to generate global features applicable for image retrieval, thereby improving localization speed.
Globally Stable Neural Imitation Policies
A. Abyaneh, Hsiu-Chin Lin
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes the SNDS method, which jointly trains the policy network and Lyapunov candidate network to formally guarantee global stability during imitation learning and achieve scalable neural policy training.
GPS-VIO Fusion with Online Rotational Calibration
Junlin Song, M. Olivares-Méndez
Autonomous DrivingSimultaneous Localization and Mapping
🎯 What it does: Proposed a GPS-VIO system capable of online calibration of rotational extrinsic parameters between the GPS and VIO reference frames
GPU-Accelerated Optimization-Based Collision Avoidance
Zeming Wu, Hao Zhang
Optimization
🎯 What it does: Proposed a GPU-accelerated optimization framework for solving collision avoidance problems where controlled objects and obstacles are modeled as convex polyhedra finite unions.
Gradient-based Local Next-best-view Planning for Improved Perception of Targeted Plant Nodes
A. K. Burusa, Gert Kootstra
OptimizationImagePoint CloudAgriculture Related
🎯 What it does: Propose a gradient-based local next best view (NBV) planning method to enhance perception accuracy for individual plant nodes in tomato greenhouses.
GrainGrasp: Dexterous Grasp Generation with Fine-grained Contact Guidance
Fuqiang Zhao, Qian Liu
OptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Proposed the GrainGrasp scheme, which uses a generative model to predict fine-grained contact maps for each finger on object point clouds, and generates precise, deterministic human-like grasping strategies based on an optimization algorithm with only point cloud input.
Graph-based 3D Collision-distance Estimation Network with Probabilistic Graph Rewiring
Minjae Song, Daehyung Park
Autonomous DrivingGraph Neural NetworkGraph
🎯 What it does: Propose a new 3D graph network, GDN-R, for data-driven collision distance estimation, achieving accurate and fast minimum distance inference through graph reconnection.
Grasp Anything: Combining Teacher-Augmented Policy Gradient Learning with Instance Segmentation to Grasp Arbitrary Objects
Malte Mosbach, Sven Behnke
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed and implemented the Teacher-Augmented Policy Gradient (TAPG) framework for grasping arbitrary objects in cluttered environments.
Grasp Manipulation Relationship Detection based on Graph Sample and Aggregation
Jiayuan Luo, Xuguang Lan
Robotic IntelligenceGraph Neural Network
🎯 What it does: Proposed the GSAGED algorithm based on graph sampling aggregation for detecting grasp operation relationships, helping robots detect targets and determine grasping order in complex scenarios.
Grasp-Anything: Large-scale Grasp Dataset from Foundation Models
An Vuong, Anh Nguyen
Data SynthesisRobotic IntelligenceLarge Language ModelImageTextMultimodality
🎯 What it does: Leverage a base model to synthesize a large-scale grasping dataset for zero-shot grasping detection.
GRF-based Predictive Flocking Control with Dynamic Pattern Formation
Chenghao Yu, Qingrui Zhang
OptimizationRobotic Intelligence
🎯 What it does: Propose a predictive swarm control algorithm based on Gibbs random fields, enabling robot swarms to autonomously follow dynamic patterns and complete shape formation in distributed environments.
Ground-Fusion: A Low-cost Ground SLAM System Robust to Corner Cases
Jie Yin, Danping Zou
Autonomous DrivingSimultaneous Localization and MappingMultimodality
🎯 What it does: Introduces a low-cost sensor fusion SLAM system called Ground-Fusion, suitable for ground vehicles, featuring efficient initialization, sensor anomaly detection and handling, real-time dense color mapping, and robust localization in diverse environments.
Grounding Conversational Robots on Vision Through Dense Captioning and Large Language Models
Lucrezia Grassi, Antonio Sgorbissa
Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelImageTextMultimodality
🎯 What it does: Combine GPT-4 with dense captioning technology to enable robots to perceive and interpret the visual world through text descriptions.
Grow Your Limits: Continuous Improvement with Real-World RL for Robotic Locomotion
Laura M. Smith, Sergey Levine
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed APRL (Policy Regularization Framework) to regulate exploration behavior in real-world robot learning, enabling quadruped robots to efficiently learn walking within minutes and continuously improve in subsequent training.
GSL-Bench: High Fidelity Gas Source Localization Benchmarking Tool
Hajo H. Erwich, G. D. Croon
BenchmarkPhysics Related
🎯 What it does: Proposes GSL-Bench as a high-fidelity benchmark tool for gas source localization (GSL), integrating realistic graphics and gas simulations to support obstacle avoidance evaluation
Guided by the Way: The Role of On-the-route Objects and Scene Text in Enhancing Outdoor Navigation
Yanjun Sun, Hirokatsu Kataoka
Autonomous DrivingVision-Language-Action ModelImageTextMultimodality
🎯 What it does: Proposes the Object-Attention VLN (OAVLN) model, helping navigation agents focus on relevant objects during training to better understand the environment and make navigation decisions.
Guided Online Distillation: Promoting Safe Reinforcement Learning by Offline Demonstration
Jinning Li, Wei Zhan
Autonomous DrivingKnowledge DistillationTransformerReinforcement LearningBenchmark
🎯 What it does: Propose a Guided Online Distillation (GOLD) framework that distills offline decision transformer (DT) policies into lightweight networks and guides exploration through online safe reinforcement learning (RL)
HabitatDyn 2.0: Dataset for Spatial Anticipation and Dynamic Object Localization
Zhengcheng Shen, Jens Lambrecht
Object DetectionImageMultimodalityBenchmark
🎯 What it does: Proposed the HabitatDyn 2.0 dataset and introduced a dynamic object localization algorithm based on spatial expectation
HAC-SLAM: Human Assisted Collaborative 3D-SLAM Through Augmented Reality
Malak Sayour, Daniel C. Asmar
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Developed a collaborative 3D SLAM system based on AR technology, integrating sensor data from mobile robots, human operators, and AR head-mounted devices to achieve automatic alignment and merging of maps, and enabling real-time editing of 3D map features through gestures on the AR interface.