arXivSub Start free trial

IROS 2024 Papers — Page 8

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

HortiBot: An Adaptive Multi-Arm System for Robotic Horticulture of Sweet Peppers

Christian Lenz, Maren Bennewitz

Robotic IntelligenceWorld ModelImageAgriculture Related

🎯 What it does: The HortiBot three-arm system integrates perception and control to complete the selective harvesting task of sweet peppers.

How Physics and Background Attributes Impact Video Transformers in Robotic Manipulation: A Case Study on Planar Pushing

Shutong Jin, Florian T. Pokorny

Robotic IntelligenceTransformerVideoPhysics Related

🎯 What it does: In robot learning, an empirical study is conducted to investigate the impact of physical attributes (color, friction coefficient, shape) and background scene features on video Transformer predictions for planar pushing trajectories, and the CloudGripper-Push-1K dataset and Video Occlusion Transformer framework are proposed.

HP3: Hierarchical Prediction-Pretrained Planning for Unprotected Left Turn

Zhihao Ou, Jian Pu

Autonomous DrivingReinforcement LearningSequential

🎯 What it does: Proposes a hierarchical predictive pre-training planning (HP3) framework for unprotected left turns, leveraging historical trajectories and complete maps to achieve general state representation and transferable scene understanding;

HPHS: Hierarchical Planning based on Hybrid Frontier Sampling for Unknown Environments Exploration

Shijun Long, Wei Fan

Robotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposed the HPHS method based on hybrid frontier sampling and hierarchical planning for rapid autonomous exploration in unknown environments; by directly sampling frontier points from LiDAR data and the robot's local map around it, and utilizing a hierarchical planning framework to divide the updated environment into several sub-regions, which are accessed in the order of total reward of the global path, reducing planning complexity and minimizing residual areas.

HS3-Bench: A Benchmark and Strong Baseline for Hyperspectral Semantic Segmentation in Driving Scenarios

Nick Theisen, Peer Neubert

SegmentationAutonomous DrivingImageBenchmark

🎯 What it does: Proposed HS3-Bench, a hyperspectral semantic segmentation benchmark for driving scenarios, and provided a strong baseline model;

HSS-SLAM: Human-in-the-Loop Semantic SLAM Represented by Superquadrics

Yulong Li, Guo-Qing Chen

Robotic IntelligenceSimultaneous Localization and MappingImage

🎯 What it does: Proposes HSS-SLAM, integrating human-computer interaction and hyper-tetrahedron object representation to achieve a semantic SLAM system;

Human Orientation Estimation Under Partial Observation

Jieting Zhao, Hong Zhang

Pose Estimation

🎯 What it does: Proposes the Part-HOE method and a confidence-aware human orientation estimation approach to handle partially visible scenarios

Human-Robot Interaction Control for Multi-Mode Exosuit with Reinforcement Learning

Kaizhen Huang, Youfu Li

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed a soft exosuit equipped with a torsion spring actuator (TSA), and proposed an adaptive impedance control method based on reinforcement learning and a nonlinear disturbance observer to address the challenge of establishing human-robot coupling dynamic models. The system optimizes impedance parameters and adjusts the robot's working mode using human motion intent, with experimental validation demonstrating its effectiveness and superiority.

Hybrid Continuum-Eversion Robot: Precise Navigation and Decontamination in Nuclear Environments using Vine Robot

Mohammed Al-Dubooni, K. Althoefer

Robotic Intelligence

🎯 What it does: Designed and constructed a hybrid continuous-deployable robot for navigation and decontamination operations in pipeline networks and enclosed remote containers within the nuclear industry, with experimental validation of its precise spraying performance.

Hybrid Stereo Dense Depth Estimation for Robotic Tasks in Industrial Automation

Suhani Singh, Jan Rosell

Depth EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Proposes a U-Net model that directly utilizes raw disparity data for dense depth reconstruction, omitting the disparity refinement step.

Hyp2Nav: Hyperbolic Planning and Curiosity for Crowd Navigation

G. M. D. D. Melendugno, Fabio Galasso

Reinforcement Learning

🎯 What it does: Proposes Hyp2Nav, a group navigation method that utilizes hyperbolic geometry learning;

Hyperbolic Image-and-Pointcloud Contrastive Learning for 3D Classification

Naiwen Hu, Jihua Zhu

ClassificationContrastive LearningImageMultimodalityPoint Cloud

🎯 What it does: Proposed a hyperbolic space-based image and point cloud contrastive learning method called HyperIPC to enhance 3D classification performance.

HyperTaxel: Hyper-Resolution for Taxel-Based Tactile Signals Through Contrastive Learning

Hongyu Li, Nawid Jamali

ClassificationPose EstimationSuper ResolutionRepresentation LearningContrastive Learning

🎯 What it does: Propose the HyperTaxel framework to learn geometric information-encoded tactile signal representations, and map sparse low-resolution trigger points to high-resolution contact surfaces via contrastive learning; enhance trigger point super-resolution by leveraging the joint probability distribution of multiple contacts; conduct qualitative and quantitative evaluations of the learned representations, and verify performance improvements in downstream tasks such as surface classification, 6D hand pose estimation, and sim-to-real transfer.

I-ASM: Iterative Acoustic Scene Mapping for Enhanced Robot Auditory Perception in Complex Indoor Environments

Linya Fu, He Kong

Robotic IntelligenceSimultaneous Localization and MappingPoint CloudSequentialAudio

🎯 What it does: Proposed an iterative framework I-ASM based on particle filtering for sound field mapping in multi-source indoor environments

I2EKF-LO: A Dual-Iteration Extended Kalman Filter Based LiDAR Odometry

Wenlu Yu, Lihua Xie

Autonomous DrivingPoint Cloud

🎯 What it does: Proposed a dual-iteration extended Kalman filter (I2EKF) and its implementation in LiDAR odometry (I2EKF-LO), which reduces motion distortion by simultaneously iterating the observation equations and state updates, and dynamically adjusts the process noise.

IC-FPS: Instance-Centroid Faster Point Sampling Framework for 3D Point-based Object Detection

Haotian Hu, Zhiwang Zhang

Object DetectionAutonomous DrivingComputational EfficiencyDiffusion modelPoint Cloud

🎯 What it does: Proposes the Instance-Centroid Faster Point Sampling (IC-FPS) framework to address the problem of low point sampling efficiency in point cloud detection, and designs the Neighboring Feature Diffusion Module (NFDM) and Centroid-Instance Sampling Strategy (CISS) to achieve fast and effective foreground discrimination and sampling.

ICR-based Kinematics for Wheeled Skid-Steer Vehicles on Firm Slopes*

Jorge L. Martínez, A. García-Cerezo

Autonomous DrivingPhysics RelatedOrdinary Differential Equation

🎯 What it does: Proposed a kinematic model for low-inertia vehicles on hard slopes based on the changes in the instantaneous center of rotation (ICR) of the side tracks during sliding steering

iDb-RRT: Sampling-based Kinodynamic Motion Planning with Motion Primitives and Trajectory Optimization

Joaquim Ortiz de Haro, Ludovic Righetti

OptimizationRobotic IntelligenceBenchmark

🎯 What it does: Propose a new sampling-based dynamic motion planning algorithm called iDb-RRT, which combines motion primitives and trajectory optimization within the RRT framework;

Identification and validation of the dynamic model of a tendon-driven anthropomorphic finger

Junnan Li, Sami Haddadin

Robotic Intelligence

🎯 What it does: A general dynamic model for tendon-driven human and humanoid fingers was developed, along with a step-by-step experimental identification and verification method; an experimental platform incorporating a drive module and peripheral sensors was designed and constructed; experiments conducted on 3D-printed Dexmart hand robotic fingers validated the established dynamic model.

Identification of Flexible Joint Robot Inertia Matrix Using Frequency Response Analysis

Kiyoung Choi, Sehoon Oh

Robotic IntelligencePhysics Related

🎯 What it does: Proposes an inertia matrix identification method for multi-degree-of-freedom flexible joint robots based on resonance and anti-resonance frequencies in the frequency response function (FRF).

Identifying Optimal Launch Sites of High-Altitude Latex-Balloons using Bayesian Optimisation for the Task of Station-Keeping

Jack Saunders, Wenbin Li

OptimizationReinforcement Learning

🎯 What it does: Study the hovering task of high-altitude latex balloons, improve the reinforcement learning controller to reduce environmental exploitation, and use Bayesian optimization to find the optimal launch location.

IDF-MFL: Infrastructure-free and Drift-free Magnetic Field Localization for Mobile Robot

Hongming Shen, Danwei Wang

OptimizationRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed an infrastructure-free, drift-agnostic mobile robot localization system based on environmental magnetic field information (IDF-MFL).

Image to Patterning: Density-specified Patterning of Micro-structured Surfaces with a Mobile Robot

Annalisa T. Taylor, Ping Guo

Robotic IntelligenceImage

🎯 What it does: Designed and implemented a credit-card-sized mobile robot that uses an adjustable tool tip to generate micro-scale pits on large surfaces, achieving surface patterns with specified density.

Image-Based Deep Reinforcement Learning with Intrinsically Motivated Stimuli: On the Execution of Complex Robotic Tasks

David Valencia, Bruce A. MacDonald

Robotic IntelligenceReinforcement LearningAuto EncoderImage

🎯 What it does: Proposed an image-based deep reinforcement learning method called NaSA-TD3 for performing complex continuous control robotic tasks

Imagine2Servo: Intelligent Visual Servoing with Diffusion-Driven Goal Generation for Robotic Tasks

Pranjali Pathre, K. M. Krishna

Robotic IntelligenceDiffusion model

🎯 What it does: Propose a visual servoing method called Imagine2Servo that utilizes diffusion models to generate intermediate target images, enabling robots to perform long-distance navigation and manipulation without predefined target images.

Imitation learning for sim-to-real adaptation of robotic cutting policies based on residual Gaussian process disturbance force model

Jamie Hathaway, Alireza Rastegarpanah

Domain AdaptationRobotic Intelligence

🎯 What it does: This paper combines Gaussian process regression with imitation learning, pairing expert actions from simulation with real observations corrected by GP to achieve sim-to-real transfer for robot cutting strategies.

Immersive Human-in-the-Loop Control: Real-Time 3D Surface Meshing and Physics Simulation

Sait Aktürk, Martin Jägersand

Robotic IntelligenceSimultaneous Localization and MappingMesh

🎯 What it does: Designed and implemented the TactiMesh Teleoperator Interface (TTI), achieving real-time 3D surface meshing and physical simulation through a head-mounted display, providing predictive visual and haptic feedback;

Implicit Neural Fusion of RGB and Far-Infrared 3D Imagery for Invisible Scenes

Xiangjie Li, Takeshi Oishi

Neural Radiance FieldImageMultimodality

🎯 What it does: Using an implicit neural fusion method (INF) to combine RGB and far-infrared images for three-dimensional reconstruction of invisible scenes.

Importance of Translational Velocity for Bird-scale Flapping Wing Vehicles Incapable of Hovering

Shijun Zhou, N. P. Hyun

Robotic IntelligencePhysics Related

🎯 What it does: Analyzes the effect of translation speed on lift gain in a 12g, 14Hz flapping, tail-equipped bird-scale flapping wing vehicle (BFWV), and proves that the vehicle cannot hover but can significantly enhance lift through forward speed.

Improved Contact Stability for Admittance Control of Industrial Robots with Inverse Model Compensation

Kangwagye Samuel, Sehoon Oh

Robotic Intelligence

🎯 What it does: A compliant controller implemented with inverse model compensation outside the position controller of an industrial robot enhances the contact stability and safety of industrial robots.

Improving behavior profile discovery for vehicles

Nelson de Moura, Fernando Garrido

Autonomous Driving

🎯 What it does: This paper proposes a novel method based on unobstructed observations at intersections to discover the behavioral characteristics of macro-maneuvers; by comparing different length trajectories of identified macro-maneuvers using the Extended Kalman Filter (EKF), and combining the Expectation-Maximization (EM) heuristic method to partition behavioral clusters; subsequently, the Kullback-Leibler divergence (KL) criterion is applied to determine the splitting and merging of clusters, ultimately dynamically and consistently identifying behavioral patterns for each macro-maneuver without relying on any map information.

Improving Legged Robot Locomotion by Quantifying Morphological Computation

V. Chandiramani, Andrew T. Conn

Robotic Intelligence

🎯 What it does: Quantified the impact of leg compliance on morphological computation (MC) in physical robotic systems and executed walking gaits on a testbed.

Improving Out-of-Distribution Generalization of Trajectory Prediction for Autonomous Driving via Polynomial Representations

Yue Yao, Joerg Reichardt

Autonomous DrivingTime SeriesSequential

🎯 What it does: Propose a unified OoD testing protocol and implement a trajectory prediction algorithm based on polynomial representation on two large-scale motion datasets, achieving near SotA ID performance and significantly improving OoD robustness.

iMTSP: Solving Min-Max Multiple Traveling Salesman Problem with Imperative Learning

Yifan Guo, Chen Wang

Optimization

🎯 What it does: Proposes an iMTSP framework based on two-layer optimization and self-supervised learning for solving the Multi-Traveling Salesman Problem (MTSP), decomposing MTSP into multiple single Traveling Salesman Problems (TSP) and achieving end-to-end learning through a longest tour self-supervised assignment network.

IMU Based Pose Reconstruction and Closed-loop Control for Soft Robotic Arms

Guanran Pei, Josie Hughes

Pose EstimationRobotic IntelligenceTime Series

🎯 What it does: Integrate multiple IMUs on the flexible robotic arm Helix to achieve pose reconstruction caused by internal and external forces, and use this dynamic pose reconstruction for kinematic-based closed-loop control;

IMU-based Monitoring of Buoy-Ballast System through Cable Dynamics Simulation

Charly Peraud, V. Hugel

Time SeriesSequentialPhysics Related

🎯 What it does: A dynamic simulation framework for cables considering variable length was constructed, and its accuracy was experimentally validated; meanwhile, the study explored improving the instrumentation scheme of the V-shaped buoy-ballast system using IMU sensors.

In-Flight Initialization of Global Visual-Inertial Estimators using Geospatial Data

Chunyu Li, Ziyang Meng

Autonomous DrivingSimultaneous Localization and MappingOptical FlowImageTabular

🎯 What it does: Using geographic information for in-air initialization of the monocular visual inertial navigation system of UAVs.

In-Hand Following of Deformable Linear Objects Using Dexterous Fingers with Tactile Sensing

Mingrui Yu, Masayoshi Tomizuka

Pose EstimationRobotic Intelligence

🎯 What it does: Implemented a technique utilizing a multi-fingered dexterous hand with tactile perception to follow deformable linear objects (DLOs) within the hand, enabling robust adjustment of grasp points while preventing object drop.

In-Hand Singulation and Scooping Manipulation with a 5 DOF Tactile Gripper

Yuhao Zhou, Yu She

ClassificationRobotic Intelligence

🎯 What it does: Designed a 5-degree-of-freedom dual-finger gripper equipped with a GelSight tactile sensor, and evaluated its performance in two tasks: retrieving, separating, and classifying objects in granular media, and grasping and accurately inserting credit cards in constrained environments.

IN-Sight: Interactive Navigation through Sight

P. Schoch, Quentin Leboutet

Autonomous DrivingRepresentation LearningRobotic IntelligenceImage

🎯 What it does: Proposed and implemented IN-Sight, a self-supervised path planning system that achieves more effective visual navigation by interacting with obstacles.

Incremental Learning of Robotic Manipulation Tasks through Virtual Reality Demonstrations

Giuseppe Rauso, Alberto Finzi

Robotic Intelligence

🎯 What it does: Proposes an incremental, modular, and scalable method for robot grasping tasks using virtual reality demonstrations with minimal samples and no prior information about objects.

Indoor Position Estimation Using NLoS Reflected Path with Wireless Distance Sensors

Tomoya Itsuka, Ryo Kurazume

OptimizationRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Simultaneously estimate the robot's position and its distance to reflective surfaces in environments with NLoS reflection paths using wireless distance sensors, addressing ranging errors caused by multipath effects.

Indoor Scene Change Understanding (SCU): Segment, Describe, and Revert Any Change

Mariia Khan, Jumana Abu-Khalaf

SegmentationRobotic IntelligenceTransformerVision Language ModelContrastive LearningImage

🎯 What it does: Propose the Scene Change Understanding (SCU) task and develop the EmbSCU method, which can simultaneously detect visual scene changes and generate robot rearrangement instructions.

Inferring Belief States in Partially-Observable Human-Robot Teams

Jack Kolb, K. Feigh

Robotic Intelligence

🎯 What it does: Explore the use of robots to infer human situational awareness through observations under limited visibility conditions.

Infrastructure-less UWB-based Active Relative Localization

Valerio Brunacci, G. Costante

Robotic IntelligenceReinforcement LearningSimultaneous Localization and Mapping

🎯 What it does: This paper proposes an active infrastructure-free UWB relative positioning method, enabling robots to actively adjust their positions to reduce the positioning error of another platform through dedicated benchmark point layouts and an adaptive loss function.

Inline Photometrically Calibrated Hybrid Visual SLAM

Nicolas Abboud, Daniel C. Asmar

Autonomous DrivingSimultaneous Localization and MappingImage

🎯 What it does: Proposed a method to integrate online sequential photometric calibration into hybrid direct-indirect visual SLAM (H-SLAM);

Insert-One: One-Shot Robust Visual-Force Servoing for Novel Object Insertion with 6-DoF Tracking

Haonan Chang, Siddarth Jain

Pose EstimationRobotic IntelligenceImage

🎯 What it does: Propose a one-shot method that achieves high-precision insertion of new objects in random poses using a single demonstration image.

Integrated 3DOF Trajectory Tracking Control for Under-actuated Marine Surface Vehicles By Trajectory Linearization

M. Sempertegui, J. J. Zhu

Autonomous Driving

🎯 What it does: Proposed and verified an integrated three-degree-of-freedom trajectory tracking control algorithm for uncontrolled marine surface vessels, adopting a multi-loop trajectory linearization control architecture, and correcting lateral drift caused by sliding turns through lateral hydrodynamic forces generated by the sideslip angle.

Integrated Electronic Circuitry for Soft Robots using Multi-Material FDM Printing

Cem Aygül, Markus P. Nemitz

Robotic Intelligence

🎯 What it does: Fabricate soft robot components and conductive paths in one step using Multi-Material Fused Deposition Modeling (FDM), and attach discrete electronic components to the conductive paths with toluene solvent to achieve electrical connections

Integrating Model-Based Footstep Planning with Model-Free Reinforcement Learning for Dynamic Legged Locomotion

Ho Jae Lee, Sangbae Kim

Robotic IntelligenceReinforcement Learning

🎯 What it does: Combine LIP model-based footstep planning with reinforcement learning (RL) strategies, training the RL policy to track desired foot positions generated by the physical model, achieving dynamic gaits without fully following the complete reference trajectory.

Integrating Online Learning and Connectivity Maintenance for Communication-Aware Multi-Robot Coordination

Yupeng Yang, Wenhao Luo

OptimizationRobotic Intelligence

🎯 What it does: Propose a data-driven control strategy that utilizes online learning and control barrier functions to maintain connectivity in multi-robot networks.

Intelligent Fish Detection System with Similarity-Aware Transformer

Shengchen Li, Zhiqiang Xu

Object DetectionTransformerVideoBenchmark

🎯 What it does: Designed a lightweight, plug-and-play edge intelligent visual system and proposed FishViT for rapid fish identification in aquatic-terrestrial transition scenarios.

Intention-Aware Planner for Robust and Safe Aerial Tracking

Qiuyu Ren, Li Xu

Object Tracking

🎯 What it does: Proposed an intent-aware planner for UAV target tracking that enhances safety and robustness when the target makes sudden movements.

Interactive Learning of Physical Object Properties Through Robot Manipulation and Database of Object Measurements

Andrej Kružliak, Matej Hoffmann

Robotic IntelligenceTabularPhysics Related

🎯 What it does: Automatically extract physical properties of objects (such as material, mass, volume, stiffness) through robot manipulation and measurement databases.

Interactive Multi-Stiffness Mixed Reality Interface: Controlling and Visualizing Robot and Environment Stiffness

Alejandro Díaz Rosales, Luka Peternel

Robotic Intelligence

🎯 What it does: Developed an interactive mixed reality interface that visualizes and controls the stiffness of a remote robot and its environment using virtual stiffness ellipsoids;

Interactive Reinforcement Learning from Natural Language Feedback

Imene Tarakli, A. D. Nuovo

Robotic IntelligenceReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Proposed the ECLAIR framework, which utilizes large language models to interpret natural language feedback and convert it into executable instructions for reinforcement learning, helping robots learn tasks more efficiently.

Interactive Reward Tuning: Interactive Visualization for Preference Elicitation

Danqing Shi, Antti Oulasvirta

Reinforcement Learning from Human FeedbackTime Series

🎯 What it does: Propose an interactive visualization-based reward weight tuning method and build a corresponding system

Interactive Robot-Environment Self-Calibration via Compliant Exploratory Actions

Podshara Chanrungmaneekul, Kaiyu Hang

Robotic Intelligence

🎯 What it does: Utilizing the robot's built-in torque sensors, the robot achieves self-calibration with the environment through its own compliance exploration actions.

Interactive-FAR:Interactive, Fast and Adaptable Routing for Navigation Among Movable Obstacles in Complex Unknown Environments

Botao He, Y. Aloimonos

Autonomous DrivingOptimization

🎯 What it does: Proposed a real-time algorithm that can actively move obstacles and dynamically adjust the global path based on sensor feedback in complex unknown environments, achieving fast and adaptive navigation.

Interpretation of Legged Locomotion in Underwater Robots based on Rimless Wheel Model

Yuetong He, Fumihiko Asano

Robotic IntelligencePhysics Related

🎯 What it does: Studied the navigation and adaptability of legged underwater robots using a rimless wheel model, and demonstrated their underwater walking and jumping behaviors through numerical simulations.

Interruptive Language Control of Bipedal Locomotion

Ashish Malik, Alan Fern

Robotic IntelligenceLarge Language ModelText

🎯 What it does: Designed and trained a bipedal robot Cassie controller capable of receiving natural language instructions at any time, and evaluated its robustness to interruptions through simulation.

IntervenGen: Interventional Data Generation for Robust and Data-Efficient Robot Imitation Learning

Ryan Hoque, Dieter Fox

Data-Centric LearningRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose the IntervenGen (I-Gen) system, which automatically generates a large amount of corrective intervention data from minimal human intervention to enhance the robustness of robot control strategies

Intraocular Reflection Modeling and Avoidance Planning in Image-Guided Ophthalmic Surgeries

Junjie Yang, I. M. A. N. Fellow

OptimizationRobotic IntelligenceImageBiomedical Data

🎯 What it does: Modeling and optimizing intraocular light reflections using microscope imaging, and integrating the optimized reflection information with path planning to avoid reflective regions, ensuring the instrument tip remains visible throughout the surgical process.

Inverse Kinematics for Neuro-Robotic Grasping with Humanoid Embodied Agents

Jan-Gerrit Habekost, Stefan Wermter

Robotic IntelligenceLarge Language ModelAgentic AI

🎯 What it does: Proposed a zero-shot motion planning method that converts Cartesian space trajectories based on Bézier curves into joint space trajectories, and applied it to two humanoid robots, NICO and NICOL, to achieve human-robot interaction grasping tasks.

Inverse Kinematics of Robotic Manipulators Using a New Learning-by-Example Method

Jacket Demby's, G. DeSouza

Robotic Intelligence

🎯 What it does: Propose an inverse kinematics method based on example learning, which takes example joint-posture pairs along with the query posture as network inputs to predict the robot's joint configuration.

Inverse Submodular Maximization with Application to Human-in-the-Loop Multi-Robot Multi-Objective Coverage Control

Guangyao Shi, Gaurav S. Sukhatme

OptimizationRobotic IntelligenceReinforcement Learning from Human Feedback

🎯 What it does: Proposed the Inverse Submodular Maximization (ISM) framework to dynamically adjust internal reward parameters in multi-robot collaboration based on suggestions from human supervisors, and developed a branch-and-bound solving algorithm tailored for this problem.

InverseMatrixVT3D: An Efficient Projection Matrix-Based Approach for 3D Occupancy Prediction

Zhenxing Ming, Stewart Worrall

Autonomous DrivingImagePoint Cloud

🎯 What it does: Propose an InverseMatrixVT3D method that efficiently converts multi-view image features into 3D feature volumes using projection matrices for 3D semantic occupancy prediction.

Investigating Behavioral and Cognitive Changes Induced by Autonomous Delivery Robots in Incidentally Copresent Persons*

Nayoung Kim, Sonya S. Kwak

Robotic IntelligenceVideoText

🎯 What it does: This study observes participants' behavioral and cognitive changes when interacting with autonomous delivery robots (ADR) in video scenarios through in-field experiments (N=30), exploring the social impact of ADR on incidental co-present individuals (InCoPs).

IR2: Implicit Rendezvous for Robotic Exploration Teams under Sparse Intermittent Connectivity

Derek Ming Siang Tan, G. Sartoretti

Robotic IntelligenceGraph Neural NetworkTransformerReinforcement Learning

🎯 What it does: Propose a deep reinforcement learning method called IR2 for information sharing in multi-robot exploration teams under sparse intermittent connectivity conditions;

Is a Simulation better than Teleoperation for Acquiring Human Manipulation Skill Data?

Donghyeon Kim, Jee-Hwan Ryu

Data SynthesisRobotic Intelligence

🎯 What it does: Compare the use of simulation and teleoperation to collect object manipulation demonstration data from human operators, and evaluate their effectiveness in tasks such as planar cutting, tight insertion, and deformable pipe coupling.

Iterative Reference Learning for Cartesian Impedance Control of Robot Manipulators

Julian M. Salt Ducaju, Rolf Johansson

Robotic Intelligence

🎯 What it does: Propose an iterative learning strategy to improve trajectory tracking of impedance-controlled robotic manipulators, propose an update rule to modify the Cartesian reference, and derive convergence conditions.

Joint Pedestrian Trajectory Prediction through Posterior Sampling

Haotian Lin, Masayoshi Tomizuka

GenerationDiffusion modelTime SeriesSequential

🎯 What it does: Proposed Guided Full Trajectory Diffuser (GFTD), a diffusion model-based framework that treats trajectory prediction as an inverse problem of space-time filling, learning complete trajectories and generating accurate and robust predictions through flexible posterior sampling, while supporting controlled generation without additional training budget.

Joint-Level IS-MPC: a Whole-Body MPC with Centroidal Feasibility for Humanoid Locomotion

Tommaso Belvedere, Giuseppe Oriolo

OptimizationRobotic Intelligence

🎯 What it does: Proposes an integrated MPC controller that utilizes full robot joint dynamics and kinematics for humanoid robot gait control.

JointLoc: A Real-time Visual Localization Framework for Planetary UAVs Based on Joint Relative and Absolute Pose Estimation

Xubo Luo, Leizheng Shu

Pose EstimationSimultaneous Localization and MappingImage

🎯 What it does: Proposed JointLoc, a real-time visual localization framework that fuses absolute 2-DoF pose and relative 6-DoF pose estimation to determine the position of planetary drones.

Jointly Learning Cost and Constraints from Demonstrations for Safe Trajectory Generation

S. Chaubey, Ville Kyrki

OptimizationRobotic IntelligenceSequential

🎯 What it does: Propose a two-step optimization process that jointly learns the cost function and constraints from demonstrations to achieve safe trajectory generation.

JUICER: Data-Efficient Imitation Learning for Robotic Assembly

Lars Ankile, Pulkit Agrawal

Data-Centric LearningRobotic IntelligenceImageTime Series

🎯 What it does: Propose an efficient imitation learning pipeline for robot assembly tasks under limited human demonstration budgets, capable of precisely grasping, reorienting, and inserting multiple components in long time sequences and multi-stage tasks;

Just Flip: Flipped Observation Generation and Optimization for Neural Radiance Fields to Cover Unobserved View

Minjae Lee, Hyeonwoo Yu

GenerationData SynthesisOptimizationNeural Radiance FieldImage

🎯 What it does: Provide data augmentation for NeRF by generating flipped observed images and estimating the camera's 6DOF pose to cover unobserved perspectives along unexplored paths of the robot.

k-Robust Conflict-Based Search with Continuous time for Multi-robot Coordination

Guilherme Daudt, Renan Maffei

OptimizationRobotic Intelligence

🎯 What it does: Proposes a multi-robot continuous-time path planning algorithm called kR-CCBS to address failures and collisions caused by navigation delays.

Kinematic Modeling of Twisted String Actuator Based on Invertible Neural Networks

Zekun Liu, Jumin Gong

Robotic IntelligenceFlow-based Model

🎯 What it does: Through a series of TSA experiments, an invertible neural network considering the number of strokes is used to establish a kinematic model for TSAs.

Kinematics-aware Trajectory Generation and Prediction with Latent Stochastic Differential Modeling

Ruochen Jiao, Qi Zhu

Autonomous DrivingAuto EncoderSequentialStochastic Differential Equation

🎯 What it does: By integrating kinematic knowledge into neural stochastic differential equations (SDEs) and designing a variational autoencoder based on the implicit kinematic-aware SDE (LK-SDE), the generation and prediction of vehicle motion are achieved.

Kinetic-energy-optimal and Safety-guaranteed Trajectory Planning for Bridge Inspection Robot Manipulator

Tianyu Zhang, Tianlong Wang

OptimizationSafty and PrivacyRobotic Intelligence

🎯 What it does: Proposes a trajectory planning method for the bridge inspection robotic arm BIRM that is both kinetic energy optimal and safety assured.

Kinodynamic Motion Planning for a Team of Multirotors Transporting a Cable-Suspended Payload in Cluttered Environments

Khaled Wahba, Wolfgang Hönig

OptimizationRobotic Intelligence

🎯 What it does: Proposes a motion planner for multi-UAV cable-based payload transportation, capable of dynamic-level planning in obstacle-dense environments.

Kiri-Spoon: A Soft Shape-Changing Utensil for Robot-Assisted Feeding

Maya N. Keely, Dylan P. Losey

Robotic Intelligence

🎯 What it does: Designed, modeled, and tested a soft deformable utensil called Kiri-Spoon aimed at improving the effectiveness of robot-assisted feeding.

KLILO: Kalman Filter based LiDAR-Inertial-Leg Odometry for Legged Robots

Shaohang Xu, Lijun Zhu

Robotic IntelligenceSimultaneous Localization and MappingPoint CloudTime Series

🎯 What it does: Designed and implemented a legged robot odometry system KLILO based on Kalman filters, fusing LiDAR, IMU, joint encoders, and contact force sensors for navigation in challenging environments.

Knowledge-based Programming by Demonstration using semantic action models for industrial assembly

Junsheng Ding, A. Perzylo

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Proposes a knowledge-based programming demonstration (kb-PbD) paradigm to enable robot programming in small and medium enterprises (SMEs).

Koopman Dynamic Modeling for Global and Unified Representations of Rigid Body Systems Making and Breaking Contact

Cormac O’Neill, H. Asada

Robotic IntelligencePhysics Related

🎯 What it does: A global modeling method based on Koopman operator theory is proposed to handle the dynamics of rigid bodies switching between contact and non-contact states, unifying various dynamic equations and eliminating the need for explicit switching; the method has been applied to dynamic peg modeling and modeling of sliding objects affected by complex friction and damping.

KOSMOS-E : Learning to Follow Instruction for Robotic Grasping

Zhi Wang, Furu Wei

Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelMultimodality

🎯 What it does: Propose KOSMOS-E, a multimodal large language model that enhances the robot's ability in fine and complex grasping tasks by using instruction-following robotic grasping data

LA-LIO: Robust Localizability-Aware LiDAR-Inertial Odometry for Challenging Scenes

Junjie Huang, Wei Liu

Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a robust locally perceptible localization LiDAR-Inertial Odometry (LA-LIO), including LiDAR degradation detection, point cloud segmentation, and adaptive-weighted error state Kalman filter (ESKF)

LAC-Net: Linear-Fusion Attention-Guided Convolutional Network for Accurate Robotic Grasping Under the Occlusion

Jinyu Zhang, Yanwei Fu

SegmentationRobotic IntelligenceConvolutional Neural NetworkMultimodality

🎯 What it does: Propose a linear fusion attention-guided convolutional network (LAC-Net), which combines visible masks generated by traditional segmentation algorithms with RGB+depth features to recover complete masks of occluded objects, thereby improving the accuracy of robotic grasping.

LANCAR: Leveraging Language for Context-Aware Robot Locomotion in Unstructured Environments

Chak Lam Shek, A. S. Bedi

Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Proposing a method that utilizes language context to enable robots to achieve context-aware mobility in unstructured environments

Lang2LTL-2: Grounding Spatiotemporal Navigation Commands Using Large Language and Vision-Language Models

J. Liu, D. Paulius

Robotic IntelligenceTransformerLarge Language ModelVision Language ModelText

🎯 What it does: Proposes the Lang2LTL-2 modular system, which leverages pre-trained large language models, vision-language models, and multi-modal semantic information to convert spatiotemporal navigation commands into structured task specifications.

Language-driven Grasp Detection with Mask-guided Attention

T. Vo, Anh Nguyen

Robotic IntelligenceTransformerVision-Language-Action ModelMultimodality

🎯 What it does: Propose a natural language guided grasp detection method combined with mask-guided attention mechanism

Language-Embedded Gaussian Splats (LEGS): Incrementally Building Room-Scale Representations with a Mobile Robot

Justin Yu, Ken Goldberg

Representation LearningRobotic IntelligenceVision Language ModelGaussian SplattingImage

🎯 What it does: This work proposes an incremental Language-Embedded Gaussian Splatting (LEGS) mapping system to build 3D scene representations that incorporate both appearance and semantic information, and support localization for open-vocabulary object queries.

Language-Guided Pattern Formation for Swarm Robotics with Multi-Agent Reinforcement Learning

Hsu-Shen Liu, Chun-Yi Lee

Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningText

🎯 What it does: Proposed a language-guided pattern formation framework LGPF based on large language models, which converts natural language descriptions into spatial pattern coordinates and enables unmanned robots to form specified patterns without collisions through CTDE multi-agent reinforcement learning;

Large Language Models Powered Context-aware Motion Prediction in Autonomous Driving

Xiaoji Zheng, Jiangtao Gong

Autonomous DrivingTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality

🎯 What it does: Using large language models (LLM) combined with visualized traffic environment (TC-Map) and text prompts to encode traffic semantic context, and integrating this contextual information into motion prediction models to improve prediction accuracy.

Large-scale Deployment of Vision-based Tactile Sensors on Multi-fingered Grippers

Meng Wang, Hangxin Liu

Robotic IntelligenceImage

🎯 What it does: Developed a low-latency synchronous image acquisition system, modular plug-and-play visual tactile sensor design, and zero-shot calibration method, and validated them on a small three-finger robotic gripper.

Large-scale Indoor Mapping with Failure Detection and Recovery in SLAM

S. Rahman, V. Kulathumani

Robotic IntelligenceSimultaneous Localization and MappingMultimodalityPoint Cloud

🎯 What it does: Proposed an automatic map generation service that combines failure detection based on visual feature tracking quality and continuous session merging, capable of handling erroneous data, scaling to large-scale indoor environments, and providing dense 3D reconstruction using depth cameras.

Latent Disentanglement for Low Light Image Enhancement

Zhihao Zheng, Mooi Choo Choo Chuah

RestorationConvolutional Neural NetworkImage

🎯 What it does: Proposes a low-light image enhancement network (LDE-Net) based on latent disentanglement, which disentangles content and illumination components in the latent space, designs a content-aware embedding (CAE) module for low-light enhancement tasks, and further develops lightweight enhancers for downstream tasks such as nighttime drone tracking and low-light object detection.

Latent Object Characteristics Recognition with Visual to Haptic-Audio Cross-modal Transfer Learning

Namiko Saito, S. Vijayakumar

RecognitionRobotic IntelligenceImageMultimodalityAudio

🎯 What it does: Proposed a cross-modal transfer learning framework from visual to tactile-audio modalities for identifying object features that are not directly observable during robotic grasping, with online recognition implemented on the Nextage Open robot.

LAVA: Long-horizon Visual Action based Food Acquisition

Amisha Bhaskar, Pratap Tokekar

Robotic Intelligence

🎯 What it does: Developed a long-term visual action primitive-based food acquisition method called LAVA, capable of sequentially acquiring liquid, semi-solid, and deformable foods from a bowl.