arXivSub Start free trial

IROS 2024 Papers — Page 7

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1581 papers

FDNet: Feature Decoupling Framework for Trajectory Prediction

Yuhang Li, Guoren Wang

Autonomous DrivingRepresentation LearningSequential

🎯 What it does: Proposed FDNet, a feature disentanglement framework for trajectory prediction

Feasibility-Guided Safety-Aware Model Predictive Control for Jump Markov Linear Systems

Zakariya Laouar, Zachary Sunberg

Optimization

🎯 What it does: A feasibility-guided, safety-oriented model predictive control (MPC) framework is designed for jump Markov linear systems (JMLS), capable of generating control strategies under uncertain mode switching and mode estimation conditions.

Feasible Region Construction by Polygon Merging for Continuous Bipedal Walking*

Chao Li, Zhihong Jiang

OptimizationRobotic IntelligenceMesh

🎯 What it does: Constructing a feasible region for continuous bipedal walking using polygon merging techniques

FEDORA: A Flying Event Dataset fOr Reactive behAvior

Amogh Joshi, Kaushik Roy

Data SynthesisImageMultimodalityTime SeriesBenchmark

🎯 What it does: Constructed a fully synthetic flight event dataset named FEDORA, containing raw data from frame cameras, event cameras, and IMU, and providing high-frequency depth, pose, and optical flow ground truth for multi-perception tasks.

FedRC: A Rapid-Converged Hierarchical Federated Learning Framework in Street Scene Semantic Understanding

Wei-Bin Kou, Yik-Chung Wu

SegmentationAutonomous DrivingFederated LearningImage

🎯 What it does: Propose a fast-converging hierarchical federated learning framework, FedRC, to address the generalization problem caused by data heterogeneity between cities in street scene semantic understanding.

Feeling Optimistic? Ambiguity Attitudes for Online Decision Making

J. Beard, Yu Gu

Optimization

🎯 What it does: Proposed and implemented the Ambiguity Attitude Graph Search (AAGS) algorithm to improve online decision-making by incorporating an adjustable ambiguity attitude when facing multiple possible transition models;

Feelit: Combining Compliant Shape Displays with Vision-Based Tactile Sensors for Real-Time Teletaction

Oscar Yu, Yu She

Robotic IntelligenceImage

🎯 What it does: Designed and implemented a low-cost remote tactile device (Feelit) that utilizes real-time information from visual-based tactile sensors, achieving tactile transmission through physical 3D surface reconstruction and shear displacement, with its performance validated in shape recognition and relative weight identification experiments.

Few-shot Transparent Instance Segmentation for Bin Picking

Anoop Cherian, Tim K. Marks

SegmentationData SynthesisRobotic IntelligenceMeta LearningConvolutional Neural NetworkImage

🎯 What it does: This paper proposes a few-shot transparent instance segmentation method called TrInSeg for robot bin picking scenarios.

FF-SRL: High Performance GPU-Based Surgical Simulation For Robot Learning

D. Dall'Alba, P. Korzeniowski

Robotic IntelligenceReinforcement LearningBiomedical Data

🎯 What it does: Proposed FF-SRL, a high-performance surgical simulation environment with a fully GPU-based implementation.

FI-SLAM: Feature Fusion and Instance Reconstruction for Neural Implicit SLAM

Xingshuo Wang, Xuanhua Chen

Neural Radiance FieldSimultaneous Localization and Mapping

🎯 What it does: Proposed and implemented a dense semantic instance SLAM system FI-SLAM based on neural implicit representations;

Fine Manipulation Using a Tactile Skin: Learning in Simulation and Sim-to-Real Transfer

Ulf Kasolowsky, Berthold Bäuml

Domain AdaptationRobotic IntelligenceReinforcement Learning

🎯 What it does: Achieving fine manipulation using tactile skin with spatial resolution on a multi-fingered robot hand through deep reinforcement learning; proposing a tactile skin model that considers fingertip softness and enables contact diffusion across multiple sensing units, with self-contained parameter calibration without external tools; learning challenging fine manipulation tasks such as rolling marbles and bolts between two fingers in a simulated environment, and demonstrating successful transfer to a real robot hand.

Fine-tuning the Diffusion Model and Distilling Informative Priors for Sparse-view 3D Reconstruction

Jiadong Tang, Mengyin Fu

GenerationKnowledge DistillationPrompt EngineeringDiffusion modelImageText

🎯 What it does: Fine-tune pre-trained diffusion models conditioned on coarse rendering results and text prompts to generate 3D-aware images, and design a semantic switcher to filter mismatched images, distilling valuable priors into 3D reconstruction models.

Finetuning Pre-trained Model with Limited Data for LiDAR-based 3D Object Detection by Bridging Domain Gaps

J. Jang, Jinkyu Kim

Object DetectionDomain AdaptationAutonomous DrivingKnowledge DistillationSupervised Fine-TuningPoint Cloud

🎯 What it does: Under limited target domain data (about 100 LiDAR frames), the Domain Adaptive Distill-Tuning (DADT) method is proposed to fine-tune pre-trained models. By employing a teacher-student architecture, object-level and context-level representations are aligned to maintain expressiveness and prevent overfitting.

Fingertip Tactile Sensor for Detecting Rope Slip

T. Koga, Shunsuke Kudoh

ClassificationRobotic IntelligenceTime Series

🎯 What it does: Developed a 3×3 tactile sensor matrix covering an elastic grip to detect rope slippage on robot fingers and created a corresponding training dataset for learning.

Fixing symbolic plans with reinforcement learning in object-based action spaces

Christopher Thierauf, Matthias Scheutz

Robotic IntelligenceReinforcement Learning

🎯 What it does: Adopted a reinforcement learning method based on an action space defined by object trajectories, replacing the traditional robot joint action space;

Flexible and Topological Consistent Local Replanning for Multirotors

Dong Wang, Fei Gao

OptimizationRobotic Intelligence

🎯 What it does: Propose a flexible and topology-consistent local replanning framework for multirotor UAVs.

Flexible Informed Trees (FIT*): Adaptive Batch-Size Approach in Informed Sampling-Based Path Planning

Liding Zhang, A. Knoll

OptimizationRobotic Intelligence

🎯 What it does: Proposes FIT*, a sampling-based path planning algorithm that flexibly adjusts batch size to enhance initial path convergence speed and later optimization efficiency.

FlexLoc: Conditional Neural Networks for Zero-Shot Sensor Perspective Invariance in Object Localization with Distributed Multimodal Sensors

Jason Wu, Mani Srivastava

Object DetectionMultimodality

🎯 What it does: Propose FlexLoc, which utilizes conditional neural networks to generate subsets of model weights at runtime based on node poses, achieving zero-shot sensor-invariant target localization;

Flight Structure Optimization of Modular Reconfigurable UAVs

Yao Su, Hangxin Liu

Optimization

🎯 What it does: Reconfigure large modular drone swarms using genetic algorithms to generate overactuated flight structures with superior dynamic performance.

FlowTrack: Point-level Flow Network for 3D Single Object Tracking

Shuo Li, Zheng Fang

Object TrackingAutonomous DrivingOptical FlowVideoPoint Cloud

🎯 What it does: Proposed a 3D single-target tracking method called FlowTrack based on point-level flow for multi-frame information

Fly by Book: How to Train a Humanoid Robot to Fly an Airplane using Large Language Models

Hyungjoo Kim, David Hyunchul Shim

Robotic IntelligenceTransformerLarge Language ModelPrompt EngineeringTextRetrieval-Augmented Generation

🎯 What it does: Training a humanoid robot (PIBOT) using a large language model to generate and execute flight mission plans in real-time based on the pilot's operating handbook (POH), completing aircraft piloting on a full-scale simulator

Flying Robotics Art: ROS-based Drone Draws the Record-Breaking Mural

Andrei A. Korigodskii, Georgii E. Bondar

Robotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Designed and deployed an autonomous drone system for creating the world's largest drone mural.

FOCWS: A High Sensitive Flexible Optical Curvature Sensor Inspired by Arthropod Sensory Systems

Jiachen Wei, Yanhong Liu

🎯 What it does: Propose a curvature sensor based on a flexible optical crack waveguide structure (FOCWS), and enhance the angular measurement sensitivity by increasing the optical power loss during bending through cutting the optical core, completing simulation and experimental verification of the optical propagation characteristics and geometric parameters.

FogROS2-FT: Fault Tolerant Cloud Robotics

†. K. Chen, Ken Goldberg

Robotic Intelligence

🎯 What it does: Built the FogROS2-FT multi-cloud fault-tolerant framework, which automatically replicates stateless robot services and routes requests to replicas, returning the first response; supports utilizing cloud computing even when cloud services are unavailable or network QoS is poor, and can safely use low-cost, easily shut-down 'spot' instances.

Foot Arch Stiffness-Based Dynamic Plantar Support Control of Human Walking Gait with Active Pneumatic Insoles

Chenhao Liu, Tao Liu

OptimizationBiomedical Data

🎯 What it does: This study proposes an active pneumatic arch support insole (PASI) combined with dynamic foot sole support control based on arch stiffness, aiming to reduce the metabolic cost of walking.

FootstepNet: an Efficient Actor-Critic Method for Fast On-line Bipedal Footstep Planning and Forecasting

Clément Gaspard, Olivier Ly

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a deep reinforcement learning (Actor-Critic) framework called FootstepNet based on a continuous action space, for rapid online bipedal gait planning in obstacle-filled local environments, with footstep count prediction functionality.

Force and Velocity Prediction in Human-Robot Collaborative Transportation Tasks through Video Retentive Networks

J. E. Domínguez-Vidal, Alberto Sanfeliu

Robotic IntelligenceVideo

🎯 What it does: Studied force/speed prediction for human-robot collaborative object transportation tasks, and extended Retentive Networks to an architecture capable of processing video inputs.

Force-Triggered Control Design for User Intent-Driven Assistive Upper-Limb Robots

Maxime Manzano, Marie Babel

Robotic Intelligence

🎯 What it does: Proposed a force-triggered (FT) controller for upper limb exoskeletons and conducted pre-experiments in pick-and-place tasks;

Formalization of Temporal and Spatial Constraints of Bimanual Manipulation Categories

F. Krebs, Tamim Asfour

🎯 What it does: Based on a bimanual operation taxonomy, a formal description of spatiotemporal constraints for each category was proposed, which was embedded into category-specific controllers to achieve real-time adaptation of bimanual collaborative actions.

Four-Axis Adaptive Fingers Hand for Object Insertion: FAAF Hand

Naoki Fukaya, Kuniyuki Takahashi

Robotic Intelligence

🎯 What it does: Designed and implemented a four-axis adaptive gripper (FAAF Hand) capable of passively adapting in the x, y, z axes, and yaw direction to complete object insertion tasks under positioning errors.

FoveaCam++: Systems-Level Advances for Long Range Multi-Object High-Resolution Tracking

Yuxuan Zhang, Sanjeev J. Koppal

Object TrackingVideo

🎯 What it does: Propose a dual-camera system to achieve high-resolution multi-target tracking up to nearly 1 km and construct a real-time image fusion pipeline;

Fractional Order Modeling and Control of Hydrogel-based Soft Pneumatic Bending Actuators

Jesús De La Morena, Andrés S. Vázquez

Physics RelatedOrdinary Differential Equation

🎯 What it does: This paper identifies the dynamic behavior of soft pneumatic bending actuators (SPBA) through experimental data and models it as a fractional-order model (FOM).

FRAGG-Map: Frustum Accelerated GPU-Based Grid Map

Michele Grimaldi, Pere Ridao Rodriguez

Autonomous DrivingComputational EfficiencySimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposed a new GPU-accelerated voxel map called FRAGG-Map, which maintains real-time updates even after multiple expansions and efficiently inserts point cloud data through CUDA kernels.

From CAD to URDF: Co-Design of a Jet-Powered Humanoid Robot Including CAD Geometry

Punith Reddy Vanteddu, Daniele Pucci

OptimizationRobotic IntelligenceMesh

🎯 What it does: Proposes a joint optimization framework for simultaneously improving the mechanical design and control performance of the body;

From LLMs to Actions: Latent Codes as Bridges in Hierarchical Robot Control

Yide Shentu, Pieter Abbeel

Robotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Proposes an architecture (LCB) that uses a learnable latent code to bridge the gap between LLM and low-level control strategies, mitigating the limitations of pure language interfaces;

Frontier-Based Exploration for Multi-Robot Rendezvous in Communication-Restricted Unknown Environments

Mauro Tellaroli, Nicola Basilico

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposes a strategy that unifies frontier exploration with multi-robot rendezvous, introducing a mechanism allowing robots to revisit explored areas to enhance rendezvous opportunities.

Frozen Assets: Leveraging Ice, Water, and Phase Transitions in Robots

Aaron Wilhelm, E. F. Helbling

Robotic Intelligence

🎯 What it does: Demonstrates how the solid and liquid states of water and their phase transitions can be utilized in common robot designs such as modular robots, robotic arms, vehicle-mounted robots, and soft robots.

FruitNeRF: A Unified Neural Radiance Field based Fruit Counting Framework

Lukas Meyer, Marc Stamminger

Object DetectionSegmentationNeural Radiance FieldImagePoint CloudAgriculture Related

🎯 What it does: We propose the FruitNeRF framework, which directly counts any fruits in 3D space by training a semantic neural radiance field;

Functional kinematic and kinetic requirements of the upper limb during activities of daily living: a recommendation on necessary joint capabilities for prosthetic arms

Christopher Herneth, Sami Haddadin

OptimizationRobotic IntelligenceVideoBiomedical Data

🎯 What it does: Analyzed motion capture data of 12 subjects performing 24 activities of daily living (ADL), calculating the range of motion, velocity, and torque of the shoulder, elbow, radioulnar, and wrist joints; used a linear regression model to predict joint torques under different limb and object masses; optimized the axial configuration of two series and two differential joint configurations of the wrist based on a data-driven approach, achieving reduced peak power.

FusionTrack: An Online 3D Multi-object Tracking Framework Based on Camera-LiDAR Fusion

Weizhen Zeng, Bingzhao Gao

Object TrackingAutonomous DrivingImageMultimodalityPoint CloudBenchmark

🎯 What it does: Proposed an online 3D multi-object tracking framework based on camera-radar fusion, primarily relying on 3D object motion models while fully utilizing image appearance information and 2D detection results.

Gaining the Sparse Rewards by Exploring Lottery Tickets in Spiking Neural Networks

Hao Cheng, Renjing Xu

Computational EfficiencySpiking Neural NetworkTransformer

🎯 What it does: Studied Spike-based Lottery Tickets (SLTs), explored their properties, and achieved two significant sparse rewards through experiments; meanwhile, proposed a sparse algorithm for the Spike Transformer structure, integrating convolutional operations into the Patch Embedding Projection (ConvPEP) module to achieve multi-level sparsification.

GazeMotion: Gaze-guided Human Motion Forecasting

Zhiming Hu, Andreas Bulling

Pose EstimationGraph Neural NetworkMultimodalityBenchmark

🎯 What it does: Proposed the GazeMotion method, which utilizes past eye gaze information and human posture to predict future human motion.

GDM-Net: Gas Distribution Mapping with a Mobile Robot Using Deep Reinforcement Learning and Gaussian Process Regression

Iliya Kulbaka, Swapnoneel Roy

Robotic IntelligenceReinforcement LearningPhysics Related

🎯 What it does: Proposed and implemented a technique combining deep reinforcement learning and Gaussian process regression for gas distribution mapping by mobile robots under limited travel budgets.

GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators

Philipp Wu, P. Abbeel

Robotic Intelligence

🎯 What it does: Built a low-cost, user-friendly, and intuitive teleoperation framework called GELLO for demonstration collection in robotic arm operations, implemented on three common robots (Franka, UR5, xArm).

GenCHiP: Generating Robot Policy Code for High-Precision and Contact-Rich Manipulation Tasks

Kaylee Burns, Karol Hausman

Robotic IntelligenceLarge Language ModelBenchmark

🎯 What it does: Generate robot strategy code using large language models (LLMs), and verify its effectiveness in high-precision and contact-rich manipulation tasks by reparameterizing the action space to include constraints on interaction forces and stiffness.

Generating Continuous Paths On Learned Constraint Manifolds Using Policy Search

Ethan Canzini, Ashutosh Tiwari

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Train a single learning manifold, incrementally determine constraint manifolds under different constraints using it, and compute the geodesic between two points via policy search to obtain the shortest path.

Generating Force Vectors from Projective Truncated Signed Distance Fields for Collision Avoidance and Haptic Feedback

Seongjin Bien, Sami Haddadin

Depth EstimationRobotic Intelligence

🎯 What it does: Propose a unified system that combines projective truncated signed distance fields (TSDF) with Euclidean signed distance fields (ESDF), generating force vectors near obstacles from TSDF reconstruction and implementing haptic feedback teleoperation based on this.

GenerOcc: Self-supervised Framework of Real-time 3D Occupancy Prediction for Monocular Generic Cameras

Xianghui Pan, Qi Chen

Depth EstimationAutonomous DrivingImage

🎯 What it does: Propose the GenerOcc self-supervised framework to achieve real-time prediction of 3D occupancy status under monocular general cameras, and collect the FishEye Dominant dataset to verify compatibility.

Geolocation on Cartographic Maps with Multi-Modal Fusion

Mengjie Zhou, Andrew Calway

RetrievalAutonomous DrivingSimultaneous Localization and MappingImageMultimodality

🎯 What it does: Proposes upgrading the matching of ground view images with 2D maps to a 2.5D space, and achieving map localization without GPS priors through a method that learns location embeddings via multi-modal feature fusion.

Geometry-aided Underwater 3D Mapping Using Side-scan Sonar

Yiqiao Yang, Zheng Fang

Simultaneous Localization and MappingUltrasound

🎯 What it does: Proposes an underwater 3D mapping framework utilizing side-scan sonar geometry, incorporating feature extraction, binocular line search for non-continuous viewpoints, and landmark localization based on scan matching adjustment.

GeRM: A Generalist Robotic Model with Mixture-of-experts for Quadruped Robot

Wenxuan Song, Donglin Wang

Robotic IntelligenceTransformerReinforcement LearningMixture of ExpertsVision-Language-Action ModelMultimodality

🎯 What it does: Proposed GeRM, a multi-task quadruped robot universal model based on offline reinforcement learning, Transformer VLA network, and Mixture-of-Experts architecture.

GESCE: Graph-based Ergodic Search in Cluttered Environments

Burhanuddin Shirose, Matthew Travers

Autonomous DrivingOptimizationGraph

🎯 What it does: Propose a hybrid motion planning algorithm that first generates a graph of the environment's free space, then uses ergodicity as a heuristic to search the graph, thereby generating paths with low ergodicity that avoid obstacles.

GestRight: Understanding the Feasibility of Gesture-driven Tele-Operation in Human-Robot Teams

Kevin Rippy, Kasthuri Jayarajah

Robotic Intelligence

🎯 What it does: Developed GestRight, a real-time gesture-driven mobile robot remote operating system, which includes a head-mounted device capturing hand joint data and converting gestures into motion commands, with user studies evaluating its effectiveness

GMMCalib: Extrinsic Calibration of LiDAR Sensors using GMM-based Joint Registration

Ilir Tahiraj, Markus Lienkamp

Autonomous DrivingPoint Cloud

🎯 What it does: Proposed GMMCalib, which achieves automatic target benchmark external calibration for multiple LiDAR sensors using GMM-based joint point cloud registration.

GNC Design and Orbital Performance Evaluation of ISS Onboard Autonomous Free-Flying Robot Int-Ball2

Taisei Nishishita, Shinji Mitani

Robotic IntelligenceSimultaneous Localization and MappingPhysics Related

🎯 What it does: Designed the GNC (navigation, attitude control) and propulsion system of Int-Ball2, verified the performance of its 6-DOF translational and rotational motion in orbit, and compared the results of ground microgravity simulation experiments with orbital experimental results.

Goal Estimation-based Adaptive Shared Control for Brain-Machine Interfaces Remote Robot Navigation

Tomoka Muraoka, Takayuki Nagai

Robotic IntelligenceBiomedical Data

🎯 What it does: Using low-frequency, discrete, and noisy user commands generated by a brain-computer interface, the system predicts user intent through target estimation technology and fuses it with high-frequency continuous commands generated by an autonomous system to achieve remote shared control of a mobile robot.

GOMA: Proactive Embodied Cooperative Communication via Goal-Oriented Mental Alignment

Lance Ying, Tianmin Shu

Robotic IntelligenceLarge Language ModelAgentic AIText

🎯 What it does: Proposes the Goal-Oriented Mental Alignment (GOMA) framework, which achieves proactive verbal collaborative communication by planning to minimize goal-related errors in the agent's mental state.

Good Things Come in Threes: The Impact of Robot Responsiveness on Workload and Trust in Multi-User Human-Robot Collaboration

Francesco Semeraro (BAE Systems Academy for Skills and Knowledge Centre), Angelo Cangelosi (BAE Systems Academy for Skills and Knowledge Centre)

Robotic Intelligence

🎯 What it does: Designed a physical collaborative experimental scenario where two users work with a collaborative robot to complete manufacturing tasks, and investigated users' perception of task workload and trust in the robot system.

Gradient-based Regularization for Action Smoothness in Robotic Control with Reinforcement Learning

I. Lee, I-Chen Wu

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: Propose and implement Gradient-based CAPS (Grad-CAPS) to reduce jitter in robot control.

Gradual Receptive Expansion Using Vision Transformer for Online 3D Bin Packing

Minjae Kang, Songhwai Oh

OptimizationTransformerReinforcement Learning

🎯 What it does: Proposed a novel reinforcement learning algorithm called GREViT for solving the online 3D bin packing problem, utilizing visual Transformers and progressive receptive field expansion techniques;

Graph Neural Network-based Multi-agent Reinforcement Learning for Resilient Distributed Coordination of Multi-Robot Systems

Anthony Goeckner, Qi Zhu

Robotic IntelligenceGraph Neural NetworkReinforcement Learning

🎯 What it does: Propose a multi-agent reinforcement learning method based on graph neural networks (GNN) for resilient distributed coordination in multi-robot systems.

GraspContrast: Self-supervised Contrastive Learning with False Negative Elimination for 6-DoF Grasp Detection

Wenshuo Wang, M. H. Ang

Pose EstimationRepresentation LearningRobotic IntelligenceContrastive LearningImagePoint Cloud

🎯 What it does: Proposed GraspContrast, a self-supervised contrastive learning framework that leverages unlabeled RGB-D images to enhance point-level feature representations for 6-DoF grasp detection

Grasping Trajectory Optimization with Point Clouds

Yu Xiang, V. Gogate

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Proposes a grasp trajectory optimization method based on point cloud representation

Gravity-aware Grasp Generation with Implicit Grasp Mode Selection for Underactuated Hands

Tianyi Ko, Koichi Nishiwaki

Data SynthesisRobotic IntelligencePhysics Related

🎯 What it does: Proposed a data generation and learning process that utilizes gravity resistance scores to train a grasp detector, favoring enclosing grasps

Greedy Perspectives: Multi-Drone View Planning for Collaborative Perception in Cluttered Environments

Krishna Suresh, Sebastian A. Scherer

OptimizationRobotic Intelligence

🎯 What it does: Developed a multi-robot multi-actor viewpoint planner with an occlusion-aware objective for collaborative cinematography in complex environments.

Grid-based Submap Joining: An Efficient Algorithm for Simultaneously Optimizing Global Occupancy Map and Local Submap Frames

Yingyu Wang, Shoudong Huang

OptimizationComputational EfficiencySimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a grid-based subgraph stitching method that uses a pose-only Gauss-Newton algorithm to simultaneously optimize the global occupancy grid and local subgraph framework, applicable to large-scale 2D non-feature SLAM.

GRID: Scene-Graph-based Instruction-driven Robotic Task Planning

Zhe-Yuan Ni, Long Zeng

Data SynthesisRobotic IntelligenceGraph Neural NetworkTransformerLarge Language ModelTextGraph

🎯 What it does: Proposed the GRID method, which iteratively generates subtasks using Scene Graphs combined with Large Language Models (LLMs) and Graph Attention Networks (GATs) to achieve instruction-based robot task planning.

GripFlexer: Development of hybrid gripper with a novel shape that can perform in narrow spaces

Donghyun Kim, Dongwon Yun

Robotic Intelligence

🎯 What it does: Developed a compact hybrid gripper called 'GripFlexer,' experimentally verifying its ability to perform complex tasks in confined spaces, particularly challenging tasks such as twisting round door handles in disaster scenarios.

Ground-Density Clustering for Approximate Agricultural Field Segmentation

H. J. Nelson, N. Papanikolopoulos

SegmentationPoint CloudAgriculture Related

🎯 What it does: Propose a new Ground-Density Quickshift++ algorithm for rapid, approximate segmentation of plants and their stems in agricultural field point clouds.

GroupTrack: Multi-Object Tracking by Using Group Motion Patterns

Xinglong Xu, Honghai Liu

Object TrackingVideo

🎯 What it does: Propose a 2D multi-target tracker named GroupTrack that leverages group motion patterns, utilizing combinations of nearby trajectories as prior information for motion prediction and data association.

Grow-to-Shape Control of Variable Length Continuum Robots via Adaptive Visual Servoing

Abhinav Gandhi, B. Çalli

Robotic IntelligenceImage

🎯 What it does: Developed an adaptive eye-hand visual control method that enables a variable-length continuous robot in 2D environments to grow from the shortest length in a closed-loop manner to a preset shape, while maintaining the target shape trajectory.

GS-Planner: A Gaussian-Splatting-based Planning Framework for Active High-Fidelity Reconstruction

Rui Jin, Fei Gao

Robotic IntelligenceGaussian Splatting

🎯 What it does: Propose the GS-Planner framework, which uses 3D Gaussian Splatting to achieve active high-fidelity reconstruction, improves 3DGS to identify unobserved regions and online assess reconstruction quality to guide the robot; design a sampling-based active exploration strategy to enhance geometric and texture quality; adopt a quadrotor robot platform, combine 3DGS safety constraints to generate executable trajectories; validate in a highly realistic simulation environment.

GSLoc: Visual Localization with 3D Gaussian Splatting

Kazii Botashev, Stamatios Lefkimmiatis

Pose EstimationGaussian SplattingImage

🎯 What it does: Proposes a visual localization method called GSLoc based on 3D Gaussian projection, which aligns panoramic cameras using pose gradients in the rendering pipeline and improves convergence with a coarse-to-fine blur kernel strategy;

GSRM: Building Roadmaps for Query-Efficient and Near-Optimal Path Planning Using a Reaction Diffusion System

Christian Henkel, Wolfgang Honig

Autonomous DrivingOptimization

🎯 What it does: Construct a road network based on the Gray-Scott reaction-diffusion system and Delaunay triangulation to enhance the efficiency of mobile robot path planning

Guiding Reinforcement Learning with Incomplete System Dynamics

Shuyuan Wang, Lixian Zhang

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose an embedded control method that utilizes incomplete system dynamics information to guide reinforcement learning.

GV-Bench: Benchmarking Local Feature Matching for Geometric Verification of Long-term Loop Closure Detection

Jingwen Yu, Hong Zhang

Simultaneous Localization and MappingBenchmark

🎯 What it does: Proposed a geometric verification benchmark tailored for long-term loop closure detection, and evaluated six local feature matching methods under this benchmark, deeply analyzing their limitations and future directions.

HabiCrowd: A High Performance Simulator for Crowd-Aware Visual Navigation

Vuong Dinh An, Anh Nguyen

Autonomous DrivingComputational EfficiencyWorld Model

🎯 What it does: Proposed HabiCrowd, a high-performance crowd-aware visual navigation simulation platform that integrates diverse human dynamic models into photorealistic environments;

Haptic Contour Following with the Smart Suction Cup

Sebastian D. Lee, Hannah S. Stuart

Robotic IntelligenceFlow-based Model

🎯 What it does: Studied the contour tracking functionality of Smart Suction Cup, exploring the capability of flow-based tactile sensors to track different speeds and various planar contours.

Hardware-Based Time Synchronization for a Multi-Sensor System

Yueqi Wang, Liao Wu

Autonomous DrivingOptimizationImageTime Series

🎯 What it does: Proposes a hardware-based multi-sensor synchronization scheme, using a Sensor Adaptor board to provide a unified clock reference, and develops a visual-inertial synchronization method that actively controls exposure duration using an ambient light sensor, aligning camera sampling moments with actual IMU sampling moments and significantly reducing time discrepancies.

Hardware-Software Co-Design for Path Planning by Drones

Ayushi Dube, S. Vrudhula

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Design a hardware-software co-design MT+, applying the Mikami-Tabuchi algorithm to onboard path planning in UAV 3D environments, and develop a dedicated hardware accelerator CDU for parallel collision detection to reduce path planning latency.

Harmonic Mobile Manipulation

Ruihan Yang, Kiana Ehsani

Robotic IntelligenceImageBenchmark

🎯 What it does: Proposes an end-to-end learning method called HarmonicMM, which can simultaneously optimize mobility and manipulation tasks. Its effectiveness is validated in both simulation and real-world environments, demonstrating adaptability to unseen settings without additional tuning, and successfully deployed in real apartments using only RGB visual inputs.

Harnessing Natural Oscillations for High-Speed, Efficient Asymmetrical Locomotion in Quadrupedal Robots

Jing Cheng, Zhenyu Gan

Robotic IntelligenceOrdinary Differential Equation

🎯 What it does: Investigated the dynamics of quadruped robots in asymmetric bounding gait, focusing on integrating trunk pitch and hip movements to enhance speed and stability.

Harnessing Symmetry Breaking in Soft Robotics: A Novel Approach for Underactuated Fingers

Ryman Hashem, Fumiya Iida

Robotic Intelligence

🎯 What it does: Studied a control method that utilizes symmetry breaking in a soft robotic fingertip self-organization mechanism to achieve object rotation within the hand.

Harnessing with Twisting: Single-Arm Deformable Linear Object Manipulation for Industrial Harnessing Task

Xiang Zhang, Masayoshi Tomizuka

Robotic Intelligence

🎯 What it does: Proposed a single-arm wire harnessing production line, generating necessary tension through robotic twisting motions to achieve precise fixture insertion, using only a single-arm robot and an integrated force/torque sensor;

Head-Mounted Hydraulic Needle Driver for Targeted Interventions in Neurosurgery

Zhiwei Fang, Hongliang Ren

Robotic IntelligenceBiomedical DataMagnetic Resonance Imaging

🎯 What it does: Designed and manufactured a 5-DoF head-mounted hydraulic needle driver robot for deep brain needle insertion and neural imaging

Heading Control for Obstacle Avoidance using Dynamic Posture Manipulation during Tumbling Locomotion

Adarsh Salagame (Northeastern University), Alireza Ramezani (Northeastern University)

Robotic Intelligence

🎯 What it does: This paper describes the dynamics of dynamic attitude manipulation during self-rotation motion through a mathematical framework, and verifies the obstacle avoidance capability achieved solely through attitude manipulation in hardware tests.

HeLiMOS: A Dataset for Moving Object Segmentation in 3D Point Clouds From Heterogeneous LiDAR Sensors

Hyungtae Lim, C. Stachniss

SegmentationPoint Cloud

🎯 What it does: Provides a 3D point cloud moving object segmentation label dataset HeLiMOS containing four heterogeneous LiDAR sensors (including two solid-state LiDARs), and proposes an automatic annotation method.

HeteroLight: A General and Efficient Learning Approach for Heterogeneous Traffic Signal Control

Yifeng Zhang, G. Sartoretti

Autonomous DrivingOptimizationTransformerReinforcement LearningGraph

🎯 What it does: Proposes a general traffic signal control framework called HeteroLight based on multi-agent reinforcement learning, utilizing the General Feature Extraction (GFE) module and the Intersection Specifics Extraction (ISE) module to enhance the performance of parameter-sharing models across different intersections;

HGP-RL: Distributed Hierarchical Gaussian Processes for Wi-Fi-based Relative Localization in Multi-Robot Systems

Ehsan Latif, Ramviyas Parasuraman

Computational EfficiencyRobotic Intelligence

🎯 What it does: Propose a distributed relative localization method called HGP-RL based on shared Wi-Fi access points, which models RSSI using hierarchical Gaussian processes to achieve relative localization in multi-robot systems.

Hierarchical Action Chunking Transformer: Learning Temporal Multimodality from Demonstrations with Fast Imitation Behavior

J. H. Park, J. Kwon

Robotic IntelligenceTransformerReinforcement LearningMultimodality

🎯 What it does: Proposed and implemented Hierarchical Action Chunking Transformer with Vector-quantization (HACT-Vq) for learning multimodal trajectories and fine-grained actions from multi-user demonstration data.

Hierarchical Consensus-Based Multi-Agent Reinforcement Learning for Multi-Robot Cooperation Tasks

Pu Feng, Wenjun Wu

Robotic IntelligenceReinforcement LearningContrastive Learning

🎯 What it does: Proposed the Hierarchical Consensus-based Multi-Agent Reinforcement Learning (HC-MARL) framework, which utilizes contrastive learning to achieve multi-level consensus, constructing global consensus from local observations to guide the execution of multi-robot collaborative tasks.

Hierarchical Large Scale Multirobot Path (Re)Planning

Lishuo Pan, Nora Ayanian

OptimizationRobotic Intelligence

🎯 What it does: Propose a hierarchical multi-robot path re-planning framework that divides the workspace into cells and uses hierarchical planners to achieve real-time re-planning

Hierarchical Search-Based Cooperative Motion Planning

Yuchen Wu, Yong Liu

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Proposed a leaderless hierarchical search-based cooperative motion planning (SCMP) method for unmanned ground vehicles with nonholonomic Ackermann models;

High Rate Mechanical Coupling of Interacting Objects in the Context of Needle Insertion Simulation With Haptic Feedback

C. Martin, H. Courtecuisse

Computational EfficiencyRobotic Intelligence

🎯 What it does: A high-frequency update flexible coupling contact model is proposed in acupuncture simulation to real-time track needle tip deformation and improve haptic feedback.

High-Accuracy 2-D AoA Estimation Using Lightweight UWB Arrays

Yi Li, Yuan Shen

🎯 What it does: Propose a high-precision 2D azimuth estimation method using a stereo UWB array, and provide a novel phase error calibration and phase ambiguity resolution scheme based on distributed ranging.

High-Fidelity SLAM Using Gaussian Splatting with Rendering-Guided Densification and Regularized Optimization

Shuo Sun, Martin Magnusson

Pose EstimationDepth EstimationGaussian SplattingSimultaneous Localization and MappingImage

🎯 What it does: Proposed a dense RGB-D SLAM system based on 3D Gaussian Splatting, achieving accurate pose tracking and visually realistic reconstruction.

High-Frequency Capacitive Sensing for Electrohydraulic Soft Actuators

Michel R. Vogt, Robert K. Katzschmann

Robotic Intelligence

🎯 What it does: Introduce the F-HASEL motor, adding extra electrode pairs in the Peano-HASEL motor for capacitive self-sensing, demonstrating displacement estimation under high-frequency (up to 20 Hz) excitation and under external loads, proposing a multi-F-HASEL displacement estimation circuit, and enabling real-time tracking of VR user joint rotations in wearable devices.

Highly Efficient Observation Process Based on FFT Filtering for Robot Swarm Collaborative Navigation in Unknown Environments*

Chenxi Li, Bin Liang

Computational EfficiencyRobotic Intelligence

🎯 What it does: Proposed an FFT-based filtering method for extracting safe directions in collaborative path planning for robot swarms in unknown environments, as well as for compressing and reconstructing sensor observation data;

HiLMa-Res: A General Hierarchical Framework via Residual RL for Combining Quadrupedal Locomotion and Manipulation

Xiaoyu Huang, K. Sreenath

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed and implemented HiLMa-Res, a generic hierarchical framework for simultaneously achieving continuous walking and manipulation tasks on quadruped robots.

HM3D-OVON: A Dataset and Benchmark for Open-Vocabulary Object Goal Navigation

Naoki Yokoyama, Sehoon Ha

Point CloudBenchmark

🎯 What it does: Proposed the HM3D-OVON dataset and benchmark, and systematically evaluated multiple methods on it