arXivSub Start free trial

IROS 2023 Papers — Page 12

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1195 papers

Towards Autonomous Robot Navigation in Human Populated Environments Using an Universal SFM and Parametrized MPC

Enrico Fiasché, E. Malis

OptimizationRobotic Intelligence

🎯 What it does: This paper proposes a universal social force model (USFM) and combines it with parameterized nonlinear model predictive control (MPC) based on thin plate spline radial basis functions (RBF), achieving autonomous navigation for robots in crowded human environments.

Towards Connecting Control to Perception: High-Performance Whole-Body Collision Avoidance Using Control-Compatible Obstacles

Moritz Eckhoff, Sami Haddadin

Robotic IntelligenceWorld ModelMultimodality

🎯 What it does: Proposes a method that integrates high-frequency real-time self-environment collision avoidance with low-frequency multimodal high-resolution environmental perception, achieving real-time digital representation through information compression based on geometric primitives (primitive skeletons); the framework's functionality and efficiency were validated through five experiments on a 9-degree-of-freedom robotic arm.

Towards Continuous Identification of Passive Human Joint Impedance Using Physical Human-Robot Interaction System

Bilal Tout, L. Vermeiren

Robotic IntelligenceTime Series

🎯 What it does: Proposes a continuous passive human joint impedance identification method based on robot load identification and a sliding window recursive least squares algorithm, enabling real-time updates of the joint model without external sensors.

Towards Cooperative Flight Control Using Visual-Attention

Lianhao Yin, Daniela Rus

Explainability and InterpretabilityRobotic IntelligenceImageMultimodality

🎯 What it does: Propose a vision-based aerial guardian system that achieves parallel autonomy in flight control by aligning human-machine eye-tracking with the attention of an end-to-end neural control system

Towards Flexible Biolaboratory Automation: Container Taxonomy-Based, 3D-Printed Gripper Fingers*

Henning Zwirnmann, Sami Haddadin

ClassificationRobotic Intelligence

🎯 What it does: Studied the classification of liquid containers in biological laboratories, designed and manufactured a single-mode dual-nozzle 3D printed parallel manipulator gripper finger integrating rigid and flexible materials to achieve stable grasping of various laboratory containers.

Towards Full Actuation: Reconfigurable Micro Underwater Robots

Nathalie Bauschmann, R. Seifried

Robotic Intelligence

🎯 What it does: Researched and implemented a system that connects multiple small underwater robots with rotational joints in a chain configuration to achieve shape reconfigurability, enabling hovering and target inspection in confined spaces.

Towards Legged Locomotion on Steep Planetary Terrain

Giorgio Valsecchi, Marco Hutter

Robotic IntelligenceSupervised Fine-Tuning

🎯 What it does: Compare the performance of standard walking strategies with crawling gaits on sand slopes, achieve more stable slope walking by fine-tuning a state-of-the-art locomotion framework and hardware modifications (enabling ANYmal to walk with knee movement), and integrate the stability margin metric into the training process, verifying it through experiments in simulation and Mars soil simulant.

Towards MR-Safe Concentric Bellows-Based Hydrostatic Linear Actuator for a Needle Driver

Kwan Kit Lin, S. Cheng

Robotic IntelligenceMagnetic Resonance Imaging

🎯 What it does: Developed and evaluated an MR-safe coaxial bellows linear actuator for a needle driver.

Towards Packaging Unit Detection for Automated Palletizing Tasks

Markus Völk, Richard Bormann

Data SynthesisDomain AdaptationRobotic Intelligence

🎯 What it does: A packaging unit detection method for automatic stacking tasks is proposed, which is entirely trained on synthetic data and can robustly be applied to any real-world packaging units without requiring additional training or setup.

Towards Safe and Aggressive Motion Generation for Dynamic Targets Pick-and-Place

Jun Shao, Yinchun Huang

OptimizationRobotic Intelligence

🎯 What it does: Proposed a time-optimal trajectory generation framework for dynamic object grasping and placement tasks, integrated with online perception on a robotic arm platform to grasp moving objects on a conveyor belt.

Track, Stop, and Eliminate: an Algorithm to Solve Stochastic Orienteering Problems Using MCTS

Carlos Diaz Alvarenga, Stefano Carpin

Optimization

🎯 What it does: Propose an algorithm combining MCTS and BAI for solving stochastic orienteering problems with probabilistic constraints.

Traffic Incident Database with Multiple Labels Including Various Perspective Environmental Information

Shota Nishiyama, Kensho Hara

RecognitionAutonomous DrivingVideo

🎯 What it does: Established and annotated a large-scale traffic accident database named V-TIDB containing ten environmental information labels, using multi-label learning to evaluate the model's performance in accident detection.

Training-Free Attentive-Patch Selection for Visual Place Recognition

Dongshuo Zhang, S. Lam

RecognitionConvolutional Neural NetworkImage

🎯 What it does: Propose a training-agnostic two-step attention-based patch selection method for visual place recognition.

Trajectory Tracking via Multiscale Continuous Attractor Networks

Therese Joseph, Michael Milford

Autonomous DrivingHyperparameter SearchRobotic IntelligenceImagePoint Cloud

🎯 What it does: Proposed and implemented a multi-scale continuous attractor network (MCAN) to achieve trajectory tracking over a wide speed range, realized automatic parameter tuning through genetic algorithms, and open-sourced a city-scale navigation simulator applicable to any street network.

Trajectory-Based SLAM for Indoor Mobile Robots with Limited Sensing Capabilities

Yaohui Chen, Andrea Okerholm Huttlin

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Designed and implemented a 2-D indoor SLAM system based solely on robot trajectories, capable of autonomous exploration and mapping under limited sensor conditions.

TransCAR: Transformer-Based Camera-and-Radar Fusion for 3D Object Detection

Su Pang, H. Radha

Object DetectionAutonomous DrivingTransformerMultimodality

🎯 What it does: Proposes a Transformer-based Camera- and Radar fusion scheme called TransCAR for 3D object detection.

Transparent Object Tracking with Enhanced Fusion Module

Kalyan Garigapati, Haibin Ling

Object TrackingTransformerBenchmark

🎯 What it does: Proposes a novel feature fusion technique that embeds transparency information into a fixed feature space to enhance tracking accuracy for transparent objects; and designs a new tracker architecture based on this technique.

TransTouch: Learning Transparent Objects Depth Sensing Through Sparse Touches

Liuyu Bian, Rui Chen

Depth EstimationOptimizationConvolutional Neural NetworkSupervised Fine-Tuning

🎯 What it does: Proposes a method to automatically fine-tune stereo networks using sparse touch depth labels, where tactile feedback probing systems collect labels, and a designed utility function optimizes probing positions under a fixed touch budget to enhance depth perception for real objects (especially transparent objects); simultaneously introduces confidence-based regularization to prevent overfitting.

TransUPR: A Transformer-based Plug-and-Play Uncertain Point Refiner for LiDAR Point Cloud Semantic Segmentation

Zifan Yu, Fengbo Ren

SegmentationAutonomous DrivingTransformerPoint Cloud

🎯 What it does: Proposed a Transformer-based pluggable uncertainty point refiner called TransUPR, which learns to refine uncertain points in LiDAR point cloud semantic segmentation, thereby improving segmentation performance

TTC4MCP: Monocular Collision Prediction Based on Self-Supervised TTC Estimation

Changlin Li, Ming Yang

Autonomous DrivingOptical FlowImage

🎯 What it does: Propose a collision prediction method (TTC4MCP) based on monocular camera pixel-level time-to-collision (TTC) estimation

TwistSLAM++: Fusing Multiple Modalities for Accurate Dynamic Semantic SLAM

Mathieu Gonzalez, J. Royan

Object TrackingPose EstimationAutonomous DrivingSimultaneous Localization and MappingImageMultimodalityPoint Cloud

🎯 What it does: Proposed the TwistSLAM++ system, which fuses stereo images with LiDAR data and utilizes semantic information to track potentially moving objects, associating them with 3D objects detected by LiDAR to obtain pose and dimensions. Subsequently, it performs registration between consecutive object scans to refine the pose, estimates shape using object scans, and constrains map points to estimated surfaces through bundle adjustment, thereby achieving dynamic semantic SLAM.

Two-Fingered Hand with Gear-Type Synchronization Mechanism with Magnet for Improved Small and Offset Objects Grasping: F2 Hand

Naoki Fukaya, Shin-ichi Maeda

Robotic Intelligence

🎯 What it does: A dual-finger robotic hand with a gear synchronization mechanism and magnetic components was developed, capable of simultaneously handling small-sized objects and objects offset from the hand's center.

Two-stage Train Components Defect Detection Based on Prior Knowledge

Gang Peng, Zhang Deng

Anomaly DetectionConvolutional Neural Network

🎯 What it does: Developed a two-stage train component defect detection method based on prior knowledge

Two-Stage Trajectory-Tracking Control of Cable-Driven Upper-Limb Exoskeleton Robots with Series Elastic Actuators: A Simple, Accurate, and Force-Sensorless Method

Yana Shu, Xiang Li

OptimizationRobotic Intelligence

🎯 What it does: A two-stage trajectory tracking control scheme for a cable-driven upper limb exoskeleton robot with series elastic actuators was studied.

TWO: A Simple Method of Directly Closing the Loop for LiDAR Odometry

Zhuo Zhang, Mingquan Lu

Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: A simple method named TWO is proposed for direct loop closure in LiDAR odometry.

UAV-Based Trilateration System for Localization and Tracking of Radio-Tagged Flying Insects: Development and Field Evaluation

J. Pak, H. Son

Object TrackingAgriculture Related

🎯 What it does: Developed and experimentally verified a drone-based trilateration system for locating and tracking flying insects (e.g., invasive wasps) equipped with wireless tags.

Ultra-Low Inertia 6-DOF Manipulator Arm for Touching the World

K. Nishii, Yoshihiro Okumatsu

Robotic Intelligence

🎯 What it does: Developed a 6-degree-of-freedom, low-inertia robotic arm

Ultrafast Acoustic Holography with Physics-Reinforced Self-Supervised Learning for Precise Robotic Manipulation

Qingyi Lu, Song Liu

Robotic IntelligenceConvolutional Neural NetworkPhysics Related

🎯 What it does: Developed a convolutional neural network based on self-supervised learning, achieving the encoding and reconstruction of ultrafast acoustic holography through interaction with a virtual physical environment combined with energy conservation constraints, for dynamic, non-contact microrobot manipulation.

UMIRobot: An Open-{Software, Hardware} Low-Cost Robotic Manipulator for Education

M. M. Marinho, Jiawei Zhao

Robotic Intelligence

🎯 What it does: Developed a low-cost (<$100) robot education kit called UMIRobot, combined with online/hybrid courses, enabling student teams to design, 3D print, assemble, and implement control strategies for the main controller and gripper, ultimately competing in a remote operation challenge.

Uncertainty Analysis for Accurate Ground Truth Trajectories with Robotic Total Stations

Maxime Vaidis, François Pomerleau

Robotic Intelligence

🎯 What it does: Propose a Monte Carlo method based on the fusion of three robotic total stations, calculating the uncertainty of six-degree-of-freedom trajectories through point-to-point minimization, and identifying and quantifying five major noise sources.

Uncertainty-Aware Gaussian Mixture Model for UWB Time Difference of Arrival Localization in Cluttered Environments

Wenda Zhao, Angela P. Schoellig

Optimization

🎯 What it does: Propose a joint localization and noise model learning algorithm based on a two-layer optimization framework, utilizing an uncertainty-aware Gaussian Mixture Model (GMM) to improve UWB TDOA localization performance.

Uncertainty-Aware Lidar Place Recognition in Novel Environments

Keita Mason, Dimity Miller

Autonomous DrivingMixture of ExpertsSimultaneous Localization and MappingPoint CloudBenchmark

🎯 What it does: Proposed and implemented an uncertainty-aware LiDAR localization method, established a new evaluation protocol, and constructed the first comprehensive benchmark, testing the performance of five uncertainty estimation techniques on three large-scale datasets.

Underactuated MIMO Airship Control Based on Online Data-Driven Reinforcement Learning

Derek Boase, M. S. Miah

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Developed an online model-free controller based on reinforcement learning and optimal control theory for controlling underactuated airships.

Understanding the Impact of Image Quality and Distance of Objects to Object Detection Performance

Yu Hao, Yi Fang

Object DetectionCompressionConvolutional Neural NetworkImage

🎯 What it does: Studied the effects of image spatial resolution, amplitude resolution (compression), and the distance between the target and camera on detection accuracy and computational cost, and proposed the adaptive YOLOv5 model (RA-YOLO) that dynamically adjusts the scale of the feature pyramid and detection head based on input resolution.

Understanding the Influence of Robot Motion on the Experimental Processes Present in Food Science Applications

Stefan Ilić, Josie Hughes

Robotic IntelligenceTabular

🎯 What it does: Developed a robotic automation system for producing, measuring, adjusting, and cleaning milk beverages made from water and powdered milk, and studied the effects of different process parameters on beverage pH values; simultaneously identified optimal process parameters based on collected data; compared with manual operations;

Underwater and Surface Aquatic Locomotion of Soft Biomimetic Robot Based on Bending Rolled Dielectric Elastomer Actuators

Chenyu Zhang, Xiang Qian

Robotic Intelligence

🎯 What it does: This paper proposes and demonstrates a sea anemone-like soft underwater robot that achieves propulsion through the motion of undulating fins driven by curled dielectric elastomer actuators (DEA).

Ungar - A C++ Framework for Real-Time Optimal Control Using Template Metaprogramming

Flavio De Vincenti, Stelian Coros

OptimizationRobotic Intelligence

🎯 What it does: Developed an open-source C++ framework called Ungar for implementing high-dimensional optimal control problems, demonstrated in quadrupedal locomotion and multi-robot cooperative manipulation using model predictive control.

UnLoc: A Universal Localization Method for Autonomous Vehicles using LiDAR, Radar and/or Camera Input

Muhammad Ibrahim, Ajmal Saeed Mian

Autonomous DrivingConvolutional Neural NetworkTransformerImageMultimodalityPoint Cloud

🎯 What it does: Proposes UnLoc, a unified multi-sensor localization method that supports LiDAR, Camera, and RADAR as needed under all weather conditions;

Unsupervised Deformable Ultrasound Image Registration and Its Application for Vessel Segmentation

Fnu Abhimanyu, H. Choset

SegmentationGenerationData SynthesisConvolutional Neural NetworkOptical FlowImageBiomedical DataUltrasound

🎯 What it does: Developed an unsupervised deformable ultrasound image registration model called U-RAFT, and applied it to synthetic image generation for vascular segmentation.

Unsupervised OmniMVS: Efficient Omnidirectional Depth Inference via Establishing Pseudo-Stereo Supervision

Zisong Chen, Yao Zhao

Depth EstimationConvolutional Neural NetworkImage

🎯 What it does: Proposed the first unsupervised omnidirectional multi-view stereo (Omnidirectional MVS) framework based on multiple fisheye images, and designed an efficient Un-OmniMVS network.

Upper Bounds for Localization Errors in 2D Human Pose Estimation

Patrick Schlosser, Tamim Asfour

Pose EstimationImage

🎯 What it does: Proposed a neural network model that can simultaneously predict human keypoint locations and the corresponding upper bounds of localization errors.

UPPLIED: UAV Path Planning for Inspection Through Demonstration

S. S. Kannan, B. Min

Autonomous DrivingOptimizationImage

🎯 What it does: A framework named UPPLIED for UAV path planning based on demonstration trajectories is proposed, used for visual inspection of large structures. The framework generates new inspection trajectories by parsing demonstration trajectories and optimizes the positions of generated inspection points to achieve better viewpoints.

USA-Net: Unified Semantic and Affordance Representations for Robot Memory

Benjamin Bolte, Chris Paxton

Robotic IntelligenceWorld ModelMultimodality

🎯 What it does: Propose USA-Net, constructing a unified semantic and usability representation, implementing a gradient-based planner in a differentiable map, supporting open-vocabulary goal specification.

User Interactions and Negative Examples to Improve the Learning of Semantic Rules in a Cognitive Exercise Scenario

Alejandro Suárez-Hernández, Guillem Alenyà

Robotic Intelligence

🎯 What it does: Using the INPRO2 learning framework, enabling robots to infer planning actions through teacher demonstrations for learning tasks used in cognitive practice

Using Piezoceramic-Actuated Stages in Precision Long-Stroke Motion Systems: A Design Procedure

Y. Al-Rawashdeh, M. Janaideh

OptimizationPhysics Related

🎯 What it does: This paper studies the integration of piezoelectric-driven phase stages with precise positioning in precision motion systems and proposes a design procedure for multi-phase configurations; a feedforward control based on the inverse Prandtl-Ishlinskii model is used to compensate for the dynamic behavior and hysteresis of piezoelectric ceramics, and step and scanning trajectory performance evaluations are conducted on a representative system (including a coarse adjustment translational phase stage and a single-axis fine adjustment phase stage).

Using Single Demonstrations to Define Autonomous Manipulation Contact Tasks in Unstructured Environments via Object Affordances

Frank Regal, Mitch Pryor

Representation LearningRobotic Intelligence

🎯 What it does: Propose an end-to-end system that captures a single demonstration through AR HMD, computes the availability primitives (AP) representation of objects, and real-time transmits task parameters to a mobile manipulator robot for execution.

UVIO: An UWB-Aided Visual-Inertial Odometry Framework with Bias-Compensated Anchors Initialization

Giulio Delama, A. Fornasier

Simultaneous Localization and Mapping

🎯 What it does: Proposed the UVIO framework, combining UWB and VIO to achieve low-drift localization; designed a multi-step adaptive initialization process, utilizing GDOP to determine optimal waypoints for unknown UWB base station positioning; subsequently tightly integrated base station distance measurements and bias compensation information into the VIO system to eliminate localization drift.

UVSS: Unified Video Stabilization and Stitching for Surround View of Tractor-Trailer Vehicles

Chunhui Zhu, Mengyin Fu

RestorationAutonomous DrivingConvolutional Neural NetworkOptical FlowVideo

🎯 What it does: Proposed a unified video stabilization and stitching method aimed at smoothing unstable frames and aligning images from a moving camera, specifically for tractor-trailer panoramic surround systems.

V2X-Lead: LiDAR-Based End-to-End Autonomous Driving with Vehicle-to-Everything Communication Integration

Zhi-Guo Deng, Weiming Shen

Autonomous DrivingReinforcement LearningPoint Cloud

🎯 What it does: Propose an end-to-end autonomous driving method based on LiDAR called V2X-Lead, integrating vehicle-to-everything (V2X) communication, utilizing model-agnostic offline deep reinforcement learning and multi-task learning to achieve safe and efficient driving in traffic intersections without traffic signal constraints, and generalize to unseen scenarios (e.g., roundabouts).

VADER: Vector-Quantized Generative Adversarial Network for Motion Prediction

M. S. Yasar, Tariq Iqbal

GenerationTransformerGenerative Adversarial NetworkTime SeriesSequential

🎯 What it does: Propose VADER, a sequence learning algorithm utilizing vector quantization for human motion prediction

Value of Assistance for Mobile Agents

Adi Amuzig, Sarah Keren

OptimizationRobotic Intelligence

🎯 What it does: Proposes the concept of Value of Assistance (VOA) to assess the expected cost reduction achievable by providing assistance at a given execution time, and provides a computation method for VOA based on Gaussian process modeling of the robot's future uncertainty; verifies VOA's ability to predict the reduction in average cost after assistance through simulations and real robot experiments.

Value-Informed Skill Chaining for Policy Learning of Long-Horizon Tasks with Surgical Robot

Tao Huang, Qi Dou

Robotic IntelligenceReinforcement LearningBiomedical Data

🎯 What it does: Propose a value-based skill chain (ViSkill) framework to address strategy learning in long-term tasks for surgical robots

VaPr: Variable-Precision Tensors to Accelerate Robot Motion Planning

Yu-Shun Hsiao, V. Reddi

Computational EfficiencyRobotic IntelligenceBenchmark

🎯 What it does: Proposes the Variable Precision (VaPr) search optimization method, which accelerates robot motion planning by assigning appropriate floating-point precision to large-scale tensors, thereby reducing memory bandwidth bottlenecks;

VARIQuery: VAE Segment-Based Active Learning for Query Selection in Preference-Based Reinforcement Learning

Daniel Marta, Iolanda Leite

Reinforcement LearningAuto EncoderSequential

🎯 What it does: Proposes an active query selection method based on the representation of state sequences using a Variational Autoencoder (VAE) for preference-based reinforcement learning.

VDBblox: Accurate and Efficient Distance Fields for Path Planning and Mesh Reconstruction

Yi-Feng Bai, Yaonan Wang

Computational EfficiencyRobotic IntelligenceSimultaneous Localization and MappingPoint CloudMesh

🎯 What it does: Proposed the VDBblox mapping framework, which can incrementally construct Euclidean Signed Distance Fields (ESDF) from TSDF mapping and improve mesh reconstruction quality.

Vehicle Motion Forecasting Using Prior Information and Semantic-Assisted Occupancy Grid Maps

Rabbia Asghar, C. Laugier

Autonomous Driving

🎯 What it does: We represent the scene as a Dynamic Occupancy Grid Map (DOGM), associate semantic labels with occupied cells, integrate map information, and propose a vehicle behavior prediction framework that combines deep learning spatiotemporal models with probabilistic methods.

Verifiable Goal Recognition for Autonomous Driving with Occlusions

Cillian Brewitt, Stefano V. Albrecht

RecognitionAutonomous DrivingExplainability and InterpretabilityTime SeriesSequential

🎯 What it does: Propose an interpretable decision tree object recognition method called OGRIT for inferring vehicle targets in occluded environments.

VERN: Vegetation-Aware Robot Navigation in Dense Unstructured Outdoor Environments

Binghao Huang, Xiaolong Wang

Robotic IntelligenceMeta LearningPoint Cloud

🎯 What it does: Proposes a navigation method for legged robots in dense vegetation environments based on a few-shot learning classifier, 2D laser scanning, vegetation-aware navigation cost maps, and holonomic recovery behaviors.

Viewpoint Push Planning for Mapping of Unknown Confined Spaces

Nils Dengler, Maren Bennewitz

Reinforcement LearningImage

🎯 What it does: Proposes a framework based on deep reinforcement learning for viewpoint planning and push-pull action generation, aiming to map unknown confined spaces by reducing map entropy and expanding the visible space.

Vine Robot Localization Via Collision

Eugenio Frias Miranda, Laura H. Blumenschein

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed a method for localizing vine-like soft robots by utilizing collision information, using an understanding of the robot's motion and sensor feedback to predict the end-effector position in real-time.

Virtual Ski Training System that Allows Beginners to Acquire Ski Skills Based on Physical and Visual Feedbacks

Yushi Okada, Jun Ohya

Reinforcement Learning from Human FeedbackImageVideo

🎯 What it does: Developed and evaluated a VR-based skiing training system that measures speed through the center of pressure of the feet and provides first-person perspective VR images to beginners, allowing them to learn skiing skills without a real ski resort.

Vision-Based Autonomous Navigation for Unmanned Surface Vessel in Extreme Marine Conditions

Muhayyuddin Ahmed, Irfan Hussain

Object DetectionObject TrackingAutonomous DrivingConvolutional Neural NetworkGenerative Adversarial NetworkImage

🎯 What it does: Proposed and implemented a visual autonomous navigation framework for unmanned surface vessels in extreme marine conditions for target tracking

Vision-Based In-Hand Manipulation of Variously Shaped Objects via Contact Point Prediction

Yuzuka Isobe, Kazunori Umeda

Robotic IntelligenceImage

🎯 What it does: Developed an in-hand manipulation method for a two-finger parallel gripper based on image prediction of contact point changes

Vision-Based Oxy-Fuel Torch Control for Robotic Metal Cutting

James Akl, B. Çalli

Robotic IntelligenceImage

🎯 What it does: Developed a vision-based automatic metal cutting control system for oxy-fuel flame cutters, and verified its feasibility through experiments.

Vision-Based State and Pose Estimation for Robotic Bin Picking of Cables

Andrea Monguzzi, Paolo Rocco

Object DetectionPose EstimationRobotic IntelligenceImage

🎯 What it does: This paper proposes a vision-based complete workflow for detecting, classifying, estimating connector pose and status (free or occluded), and manipulating semi-variable linear objects (e.g., cables) in storage bins, using dual-arm robots to grasp, split, extract, and reassemble them; ultimately achieving the complete process of extracting a single SDLO from the storage bin to the workbench.

Vision-Based Vineyard Navigation Solution with Automatic Annotation

E. Liu, Yu Jiang

Autonomous DrivingConvolutional Neural NetworkImageAgriculture Related

🎯 What it does: Developed a vision-based automated vineyard navigation scheme that directly estimates passage heatmaps using RGB-D images, converts them into driving paths, and implements auto-annotation and row-column switching modules.

Visual Contact Pressure Estimation for Grippers in the Wild

Jeremy Collins, Charles C. Kemp

Data-Centric LearningRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Proposed and implemented a vision-based soft gripper contact pressure estimation method called ViPER, which captures images using an in-hand camera and outputs pressure distribution.

Visual Localization Based on Multiple Maps

Yukai Lin, Jiangwei Li

Pose EstimationSimultaneous Localization and MappingImage

🎯 What it does: Propose a multi-map visual localization method for image sequence localization; achieve robust and accurate camera pose estimation by constructing multiple single maps, single-image localization, consensus set maximization, and combining local SLAM for pose refinement.

Visual Pre-Training for Navigation: What Can We Learn from Noise?

Yanwei Wang, Ching-Yun Ko

Representation LearningImage

🎯 What it does: Proposed a self-supervised learning-based visual navigation pre-training method, which learns representations from synthetic noisy images by predicting random cropping positions and sizes in the target perspective, and transfers them to natural home images to achieve efficient learning of navigation strategies with minimal interaction data.

Visual Servoing on Wheels: Robust Robot Orientation Estimation in Remote Viewpoint Control

Luke Robinson, Paul Newman

Data SynthesisPose EstimationRobotic IntelligenceImage

🎯 What it does: Propose a rapidly deployable visual servoing robot process that estimates the robot's pose in 2D camera images using learning methods without assuming robot or environmental features.

Visual-Inertial-Laser-Lidar (VILL) SLAM: Real-Time Dense RGB-D Mapping for Pipe Environments

Tina Tian, Lu Li

OptimizationSimultaneous Localization and MappingMultimodalityPoint Cloud

🎯 What it does: Propose a VILL-SLAM algorithm that integrates a monocular camera, inertial sensors, a ring laser range finder, and radar for real-time dense RGB-D mapping inside pipelines.

Visual-Kinematics Graph Learning for Procedure-Agnostic Instrument Tip Segmentation in Robotic Surgeries

Jiaqi Liu, Qi Dou

SegmentationGraph Neural NetworkContrastive LearningMultimodality

🎯 What it does: Propose a visual-kinematic graph learning framework that accurately segments surgical instrument tips using multi-modal information.

Visual-LiDAR-Inertial Odometry: A New Visual-Inertial SLAM Method Based on an iPhone 12 Pro

Lingqiu Jin, Cang Ye

OptimizationSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposed a visual-LiDAR-inertial odometry (VLIO) SLAM method based on the iPhone 12 Pro.

Visual, Spatial, Geometric-Preserved Place Recognition for Cross-View and Cross-Modal Collaborative Perception

Peng Gao, Ming C. Lin

RecognitionAutonomous DrivingMultimodalityBenchmark

🎯 What it does: Proposed a cross-view cross-modal location recognition method combining semantic graph matching and distance field matching.

Visuotactile Sensor Enabled Pneumatic Device Towards Compliant Oropharyngeal Swab Sampling

Shoujie Li, Xiao-Ping Zhang

Safty and PrivacyRobotic Intelligence

🎯 What it does: Designed a low-cost, highly compliant oral-pharyngeal swab sampling device integrated with vision-tactile sensors and pneumatic actuators, enabling rapid deployment on robotic arms.

VIW-Fusion: Extrinsic Calibration and Pose Estimation for Visual-IMU-Wheel Encoder System

Chunxia Qiao, Dan Zhang

Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingImageMultimodalityTime Series

🎯 What it does: Propose a joint extrinsic calibration algorithm for the camera-IMU-wheel encoder system and a multi-sensor fusion pose estimation algorithm, while improving the VIO initialization method.

VL-Grasp: a 6-Dof Interactive Grasp Policy for Language-Oriented Objects in Cluttered Indoor Scenes

Yuhao Lu, Shengjin Wang

Pose EstimationRobotic IntelligenceVision-Language-Action ModelMultimodality

🎯 What it does: Propose a vision-and-language interactive grasping strategy, VL-Grasp, to achieve target object grasping in indoor scenes.

Volitional EMG Control Enables Stair Climbing with a Robotic Powered Knee Prosthesis

Suzi Creveling, Tommaso Lenzi

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed a proportional electromyography (EMG)-based controller enabling upper limb amputees to directly adjust the electric knee joint torque through residual leg muscle activity during stair climbing, achieving autonomous control.

Walking in Narrow Spaces: Safety-Critical Locomotion Control for Quadrupedal Robots with Duality-Based Optimization

Qiayuan Liao, K. Sreenath

OptimizationRobotic Intelligence

🎯 What it does: Propose a safety-critical quadruped robot locomotion control framework enabling the robot to navigate safely in crowded environments

WatchPed: Pedestrian Crossing Intention Prediction Using Embedded Sensors of Smartwatch

Jibran A. Abbasi, M. Won

Autonomous DrivingVideoMultimodalityTime Series

🎯 What it does: Designed, implemented, and evaluated a pedestrian crosswalk intention prediction model that combines smartwatch motion sensor data with vehicle camera visual data

Water Surface Walking of Six-Legged Robot by Controlling Attitude of Feet When It Enter Water

Yuetong He, Fumihiko Asano

Robotic Intelligence

🎯 What it does: Designed and implemented a six-legged water-walking robot composed of edgeless wheels and flywheels, and proposed a foot control method inspired by the leg movements of animals capable of walking on water. Subsequently, the control method was improved to enhance the robot's forward speed, and the impact of different physical parameters on the robot's motion was analyzed.

Weakly Supervised Caveline Detection for AUV Navigation Inside Underwater Caves

Boxiao Yu, M. Islam

Object DetectionAutonomous DrivingTransformerImage

🎯 What it does: Propose a lightweight cave line detection model based on Vision Transformer (ViT), trained using a weakly supervised manner.

Weakly Supervised Referring Expression Grounding via Dynamic Self-Knowledge Distillation

Jinpeng Mi, Jianwei Zhang

Knowledge DistillationVision Language ModelMultimodality

🎯 What it does: Propose a target-guided self-knowledge distillation framework for weakly supervised referring expression localization.

What to Learn: Features, Image Transformations, or Both?

Yuxuan Chen, T. Barfoot

Image TranslationPose EstimationConvolutional Neural NetworkImage

🎯 What it does: Propose combining an image transformation network with a feature learning network to first convert night images into daytime style and then perform feature matching, thereby enhancing long-term localization performance in visual localization.

Whole Body Control Formulation for Humanoid Robots with Closed/Parallel Kinematic Chains: Kangaroo Case Study

Sait Sovukluk, Christian Ott

Robotic Intelligence

🎯 What it does: This study extends the whole-body control (WBC) formulation for bipedal humanoid robots with closed-chain (parallel) structures, implementing it on PAL Robotics' 76-degree-of-freedom robot Kangaroo. It explores WBC formulations with two control structures: IDC and MPTC, achieving disturbance suppression of the center of mass (CoM) trajectory in 3D spring-loaded inverted pendulum (SLIP) jumping trajectories.

Whole Shape Estimation of Transparent Object from Its Contour using Statistical Shape Model

Kaihei Okada, Tetsuyou Watanabe

Image

🎯 What it does: Propose a method that utilizes statistical shape models to estimate the 3D shape of transparent objects from RGB-D images by fitting contours extracted from RGB images to determine object shape, and using depth maps to detect the plane where the object resides.

Whole-Body Torque Control Without Joint Position Control Using Vibration-Suppressed Friction Compensation for Bipedal Locomotion of Gear-Driven Torque Sensorless Humanoid

Takuma Hiraoka, Koji Kawasaki

Robotic Intelligence

🎯 What it does: Proposes a four-layer hierarchical whole-body torque control method, enabling a humanoid robot with high reduction ratio joints to perform walking and carrying tasks without torque sensors and joint position control

Wireless Capacitive Tactile Sensor Arrays for Sensitive/Delicate Robot Grasping

S. Ergun, Hubert Zangl

Robotic Intelligence

🎯 What it does: Designed and implemented a customizable wireless capacitive tactile sensor array (CTSA) for robotic grasping of sensitive/delicate objects, predicting grasp quality through low-force initial grasping to avoid damaging objects.

Wireless Network Demands of Data Products from Small Uncrewed Aerial Systems at Hurricane Ian

Thomas Manzini, Justin Adams

🎯 What it does: Collected and analyzed the demand for wireless communication networks by small unmanned aerial vehicles (UAVs) during disaster response in the context of Hurricane Ian in 2022, quantified communication bottlenecks, and proposed a data-to-decision workflow model.

WIT-UAS: A Wildland-Fire Infrared Thermal Dataset to Detect Crew Assets from Aerial Views

Andrew Jong, S. Scherer

Object DetectionData-Centric LearningImageBenchmark

🎯 What it does: Proposed the WIT-UAS dataset, which collects drone flight data and manually annotated images captured using long-wave infrared thermal imaging technology in controlled wildfire environments, for detecting personnel and vehicle assets in fire scenes.

Workspace Force/Acceleration Disturbance Observer for Precise and Safe Motion Control

Wooseok Han, Sehoon Oh

Robotic Intelligence

🎯 What it does: Proposed a new workspace force/acceleration disturbance observer (WFADOB) controller, which designs a disturbance observer using interaction force and acceleration to achieve precise motion tracking and low-impedance safe contact.

Wrench Estimation of Modular Manipulator with External Actuation and Joint Locking

Yonghyeok Kim, Dongjun Lee

Robotic Intelligence

🎯 What it does: Proposed an external torque estimation method for modular robotic hands under external actuation and joint locking conditions, designed a distributed algorithm based on recursive Newton-Euler dynamics and a centralized algorithm based on the D'Alembert principle, and validated their effectiveness through experiments.

WSCFER: Improving Facial Expression Representations by Weak Supervised Contrastive Learning

Wei Nie, Honghai Liu

RecognitionRepresentation LearningContrastive LearningImage

🎯 What it does: Propose a weakly supervised contrastive learning FER (WSCFER) method, which simultaneously learns category-level and instance-level representations by integrating three components (FER classification task, weakly supervised contrastive learning task, partial consistency loss) to enhance facial expression recognition performance.

Zero-Shot Fault Detection for Manipulators Through Bayesian Inverse Reinforcement Learning

Hanqing Zhao, Gregory Dudek

Anomaly DetectionRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a zero-shot external fault detection method based on AVRIL, which recovers reward signals from normal operation samples to detect unseen faults.

ZMP Feedback Balance Control of Humanoid in Response to Ground Acceleration

Masanori Konishi, Koji Kawasaki

Robotic Intelligence

🎯 What it does: Proposed introducing a supporting foot acceleration term into the walking steady-state controller to address balance issues caused by ground acceleration