ICRA 2023 Papers — Page 5
IEEE International Conference on Robotics and Automation · 1341 papers
Domain Generalised Fully Convolutional One Stage Detection
Karthik Seemakurthy, Charles Fox
Object DetectionDomain AdaptationConvolutional Neural NetworkImage
🎯 What it does: Propose a domain generalization method DGFCOS for the single-stage anchor-free detector FCOS to enhance detection performance on unknown target domains, and evaluate it on four datasets.
Domain-specific languages for kinematic chains and their solver algorithms: lessons learned for composable models
Sven Schneider, H. Bruyninckx
Robotic IntelligenceText
🎯 What it does: Analyzes the structure and semantics of URDF and COLLADA, two domain-specific languages (DSLs) used to represent robot kinematic chains. It points out their tight coupling with multiple domains (such as dynamics, visualization, control), and designs a composable, loosely coupled intermediate exchange format along with a complete metamodel to explicitly define model semantics. On this basis, a solver algorithm is constructed and a structure and semantics-preserving model-to-code conversion is implemented.
DOTIE - Detecting Objects through Temporal Isolation of Events using a Spiking Architecture
M. Nagaraj, Kaushik Roy
Object DetectionSpiking Neural Network
🎯 What it does: Leveraging the temporal information from event cameras, a lightweight spiking neural network is proposed for detecting moving objects
DQN-based on-line Path Planning Method for Automatic Navigation of Miniature Robots
Jialin Jiang, Li Zhang
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed an online path planning strategy based on Deep Q-Network (DQN) for real-time navigation of micro-magnetic robots at different scales.
DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learning
I. Made, Hyunsam Myung
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a new quadruped robot locomotion learning framework that achieves robust walking through implicit terrain imagination under limited perception, verified effective through a single long-distance test in outdoor variable environments.
DribbleBot: Dynamic Legged Manipulation in the Wild
Yandong Ji, Pulkit Agrawal
Domain AdaptationRobotic IntelligenceWorld ModelImage
🎯 What it does: A multi-legged robot named DribbleBot capable of dribbling a soccer ball in a real environment comparable to human conditions.
DriveIRL: Drive in Real Life with Inverse Reinforcement Learning
Tung Phan-Minh, Eric M. Wolff
Autonomous DrivingReinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: Proposed DriveIRL, the first planner using inverse reinforcement learning (IRL) for driving in dense urban traffic, which generates diverse trajectory candidates, scores them with a learned model, and tracks the best trajectory through a low-level controller.
DS-K3DOM: 3-D Dynamic Occupancy Mapping with Kernel Inference and Dempster-Shafer Evidential Theory
Ju Han, Han-Lim Choi
Simultaneous Localization and Mapping
🎯 What it does: Propose a 3-D dynamic occupancy mapping algorithm DS-K3DOM, which performs sequential updates on measurement streams using Bayesian methods based on the theory of random finite sets, and realizes real-time computation through particle approximation in the Dempster-Shafer domain. Furthermore, dense mapping from sparse measurements is achieved by employing kernel-based reasoning and Dirichlet basic belief allocation.
DTact: A Vision-Based Tactile Sensor that Measures High-Resolution 3D Geometry Directly from Darkness
Chan-Yu Lin, Huazhe Xu
Pose EstimationDepth EstimationImage
🎯 What it does: Designed and manufactured a vision-based tactile sensor DTact, leveraging the reflective properties of translucent elastomers to directly measure high-resolution 3D geometry from the shadow in captured tactile images, and perform calibration using only a single image; subsequently, a prototype with non-planar contact surfaces was fabricated and tested in two robotic tasks: pose estimation and object recognition.
Dual quaternion based dynamic movement primitives to learn industrial tasks using teleoperation
Rohit Chandra, Y. Mezouar
Robotic Intelligence
🎯 What it does: Learning manipulation of deformable objects in industrial tasks using dynamic movement primitives based on dual quaternions, validated through dual-arm robots and motion capture systems.
Dual Robot Collaborative System for Autonomous Venous Access Based on Ultrasound and Bioimpedance Sensing Technology
Maria Koskinopoulou, L. Mattos
Robotic IntelligenceImageBiomedical DataUltrasound
🎯 What it does: Designed and implemented a dual-arm collaborative autonomous central venous catheterization system, utilizing ultrasound and bioimpedance sensing technologies to achieve precise needle insertion.
DuEqNet: Dual-Equivariance Network in Outdoor 3D Object Detection for Autonomous Driving
Xihao Wang, Xian Wei
Autonomous DrivingConvolutional Neural NetworkGraph Neural NetworkPoint Cloud
🎯 What it does: Propose a Dual Equivariant Network (DuEqNet) to address the issue of decreased 3D object detection accuracy caused by vehicle steering in autonomous driving scenarios.
Dynamic Control Barrier Function-based Model Predictive Control to Safety-Critical Obstacle-Avoidance of Mobile Robot
Zhu Jian, Bin Liang
Robotic IntelligencePoint Cloud
🎯 What it does: A safety avoidance method for dynamic obstacles in LiDAR-based mobile robots, which uses point clouds to generate real-time local grid maps, clusters obstacles with DBSCAN and applies minimum bounding ellipse (MBE) closure, estimates/predicts obstacle motion trajectories using data association and Kalman filtering, parameterizes trajectories as a set of ellipses, and achieves safe dynamic obstacle avoidance by combining extended dynamic control barrier functions (D-CBF) with model predictive control (MPC).
Dynamic Interactive Relation Capturing via Scene Graph Learning for Robotic Surgical Report Generation
Hongqiu Wang, Lei Zhu
GenerationRobotic IntelligenceGraph Neural NetworkTransformerTextBiomedical Data
🎯 What it does: Propose a neural network that explicitly explores the interaction relationships between tissues and surgical instruments through scene graph learning to enhance robotic surgery report generation.
Dynamic Locomotion of a Quadruped Robot with Active Spine via Model Predictive Control
Wanyue Li, Hui Cheng
OptimizationRobotic Intelligence
🎯 What it does: Proposes an MPC-based SRB model control method considering active spinal motion, and develops a quadruped robot Yat-sen Lion with a 3-DOF active spine; experiments verify its ability to achieve bending, arching, and turning behaviors during locomotion;
Dynamic Modeling and Identification of a Robotic Intracardiac Echo Catheter
Mohammad Salehizadeh, J. Jagadeesan
Robotic IntelligenceUltrasound
🎯 What it does: Developed a dynamic model for intracardiac echocardiography (ICE) catheters as the first step toward enabling autonomous robotic ICE catheter tracking of ablation catheters and monitoring of trauma.
Dynamical System-based Imitation Learning for Visual Servoing using the Large Projection Formulation
Antonio Paolillo, Matteo Saveriano
Robotic Intelligence
🎯 What it does: Proposed a dynamical system-based imitation learning method for visual servoing, employing a large projection task priority formula.
DytanVO: Joint Refinement of Visual Odometry and Motion Segmentation in Dynamic Environments
Shihao Shen, S. Scherer
SegmentationPose EstimationSimultaneous Localization and Mapping
🎯 What it does: Propose DytanVO, a supervised learning visual odometry method that can real-time predict camera motion in dynamic environments
E-VFIA: Event-Based Video Frame Interpolation with Attention
Onur Selim Kilicc, Aydin Alatan
RestorationGenerationConvolutional Neural NetworkTransformer
🎯 What it does: Proposed a lightweight convolution-based event-driven video frame interpolation method called E-VFIA.
Edge Grasp Network: A Graph-Based SE(3)-invariant Approach to Grasp Detection
Hao-zhe Huang, Robert W. Platt
Pose EstimationRobotic IntelligenceGraph Neural NetworkPoint Cloud
🎯 What it does: Proposes a graph-based SE(3)-invariant neural network model for detecting 6-DoF grasp poses from single-view point clouds, significantly improving grasp success rates.
Edge-guided Multi-domain RGB-to-TIR image Translation for Training Vision Tasks with Challenging Labels
Dong-Guw Lee, Ayoung Kim
Image TranslationObject DetectionConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Propose an edge-guided multi-domain RGB-to-TIR image translation model, and use the generated realistic TIR images to train TIR tasks such as optical flow estimation and object detection.
EdgeVO: An Efficient and Accurate Edge-based Visual Odometry
H Zhao, F. Gu
Pose EstimationComputational EfficiencySimultaneous Localization and MappingVideo
🎯 What it does: Proposes an edge-based visual odometry method called EdgeVO.
EDO-Net: Learning Elastic Properties of Deformable Objects from Graph Dynamics
A. Longhini, Danica Kragic
Representation LearningGraph Neural NetworkGraphPhysics Related
🎯 What it does: Proposes the EDO-Net model, which learns graph dynamics of deformable objects, extracts latent representations of elastic physical properties from stretching interactions, and uses these representations to predict future states of deformable objects such as fabric.
Effect of the Dynamics of a Horizontally Wobbling Mass on Biped Walking Performance
Tomoya Kamimura, Akihito Sano
OptimizationRobotic IntelligencePhysics Related
🎯 What it does: Studied a two-legged robot model with a lateral swinging mass connected to the upper body, and explored the impact of lateral dynamics on walking by numerically searching for the model's limit cycles.
Effective Combination of Vertical, Longitudinal and Lateral Data for Vehicle Mass Estimation
Younesse El Mrhasli, X. Mouton
Autonomous DrivingOptimizationTime Series
🎯 What it does: Proposed a vehicle mass estimation method that integrates vertical, longitudinal, and lateral data
Efficient and Hybrid Decoder for Local Map Construction in Bird'-Eye-View
Kun Tian, Guan Huang
Object DetectionSegmentationAutonomous DrivingConvolutional Neural NetworkSimultaneous Localization and Mapping
🎯 What it does: Propose an efficient hybrid decoder EHD for local map construction in bird's-eye view (BEV), integrating a CNN segmentation head and a query-based lane detection head;
Efficient Bimanual Handover and Rearrangement via Symmetry-Aware Actor-Critic Learning
Yunfei Li, Yi Wu
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a symmetric perception actor-critic framework for dual-arm collaborative control in multi-object rearrangement tasks, achieving rearrangement of up to 8 objects with a success rate exceeding 70% in simulation, followed by deployment and demonstration of human-robot collaboration on two Franka Panda robotic arms.
Efficient Bundle Adjustment for Coplanar Points and Lines
Lipu Zhou, M. Kaess
Pose EstimationOptimization
🎯 What it does: Proposed a π-Bundle Adjustment (π-BA) method targeting planar coplanar points and lines, introducing new π constraints and π factors to replace traditional reprojection error;
Efficient Implicit Neural Reconstruction Using LiDAR
Dongyu Yan, Yi Lin
Autonomous DrivingOptimizationComputational EfficiencyRepresentation LearningNeural Radiance FieldPoint Cloud
🎯 What it does: Reconstruct a fine-grained implicit occupancy field within minutes using sparse LiDAR point clouds and rough odometry.
Efficient Inference of Temporal Task Specifications from Human Demonstrations using Experiment Design
Shlok Sobti, L. Kavraki
OptimizationComputational EfficiencyReinforcement Learning from Human FeedbackSequential
🎯 What it does: Using experimental design to select environments for human demonstrations to efficiently infer task specifications based on temporal logic.
Efficient Learning of High Level Plans from Play
N'uria Armengol Urp'i, Stelian Coros
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose the ELF-P framework, which learns a discrete behavior prior using task-agnostic play data and designs a high-level goal-conditioned policy to improve learning efficiency for long-horizon complex manipulation tasks.
Efficient Learning of Locomotion Skills through the Discovery of Diverse Environmental Trajectory Generator Priors
Shikha Surana, Antoine Cully
Robotic IntelligenceReinforcement Learning
🎯 What it does: Learn diverse specialized trajectory generators using the EETG method, and enable quadruped robots to navigate various terrains (ramps, stairs, rough terrain, balance beams) with a single policy in the Policies Modulating TG (PMTG) architecture.
Efficient Optimal Planning in non-FIFO Time-Dependent Flow Fields
James Ju Heon Lee, R. Fitch
Optimization
🎯 What it does: Proposes an algorithm for solving the shortest path problem in non-FIFO time-dependent flow fields.
Efficient Planar Pose Estimation via UWB Measurements
Haodong Jiang, Jun-Yi Wu
Pose Estimation
🎯 What it does: The study achieves planar pose estimation using only UWB ranging and proposes a two-step Gauss-Newton iteration method to improve estimation accuracy
Efficient Planning of Multi-Robot Collective Transport using Graph Reinforcement Learning with Higher Order Topological Abstraction
Steve Paul, Souma Chowdhury
OptimizationComputational EfficiencyRobotic IntelligenceGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: Proposed a TD-enhanced graph neural network strategy for multi-robot collective transportation tasks, improving transferability and computational speed for large-scale unknown problems.
Efficient Preference-Based Reinforcement Learning Using Learned Dynamics Models
Yi Liu, Daniel S. Brown
Reinforcement Learning from Human FeedbackReinforcement LearningWorld Model
🎯 What it does: Investigated and verified the advantages and challenges of using learned dynamics models in preference-based reinforcement learning.
Efficient Recovery Learning using Model Predictive Meta-Reasoning
Shivam Vats, Oliver Kroemer
Robotic IntelligenceMeta LearningReinforcement Learning
🎯 What it does: Propose a general method to progressively enhance operational robustness by first exploring the failure modes of the current policy in simulation and then learning additional recovery skills to handle these failures.
Efficient View Path Planning for Autonomous Implicit Reconstruction
Jing Zeng, Qianru Ye
Autonomous DrivingOptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Construct an information gain field using neural networks and combine fine-grained implicit representations with coarse voxel representations to achieve efficient viewpoint path planning.
Efficient Visual-Inertial Navigation with Point-Plane Map
Jiaxin Hu, Guoquan Huang
Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposes a large-scale visual inertial odometry system PPM-VIO based on filtering, which corrects cumulative drift by utilizing point-plane prior maps.
Efficiently Approaching Groups of People in a Socially Acceptable Manner in Environments with Obstacles
Aline F. F. Silva, D. Macharet
Optimization
🎯 What it does: Proposed a method for planning continuous access to different groups in crowded environments in a socially acceptable manner
Efficiently Learning Small Policies for Locomotion and Manipulation
Shashank Hegde (University of Southern California), G. Sukhatme (University of Southern California)
Computational EfficiencyRobotic IntelligenceGraph Neural NetworkReinforcement Learning
🎯 What it does: Propose utilizing Graph Hyper Networks to learn compact control policies applicable to memory-constrained agile robots;
EFTrack: A Lightweight Siamese Network for Aerial Object Tracking
Wenqi Zhang, Gang Yang
Object TrackingConvolutional Neural Network
🎯 What it does: Proposed a lightweight aerial object tracker based on a Siamese network, using EfficientNet as the backbone network. After pixel-level association, classification and regression branches were added to predict the positive/negative scores and offsets of the target, eliminating the need for predefined anchors.
EgoHMR: Egocentric Human Mesh Recovery via Hierarchical Latent Diffusion Model
Yuxuan Liu, Guang-Zhong Yang
Pose EstimationDiffusion modelImageMesh
🎯 What it does: We propose EgoHMR, a hierarchical network based on latent diffusion models, which can recover human meshes from single-frame first-person images and can be trained end-to-end without requiring 2D joint supervision.
Elastic Context: Encoding Elasticity for Data-driven Models of Textiles Elastic Context: Encoding Elasticity for Data-driven Models of Textiles
A. Longhini, D. Kragic
Data-Centric LearningRobotic Intelligence
🎯 What it does: Proposed the Elastic Context (EC) method, which encodes the elasticity of textiles through stress-strain curves and applies it to robot interaction;
Electroadhesive Auxetics as Programmable Layer Jamming Skins for Formable Crust Shape Displays
Ahad M. Rauf, Sean Follmer
Robotic IntelligencePhysics Related
🎯 What it does: Proposed and implemented the use of electroactive auxetic materials as a deformation constraint layer for programmable deformation in continuous shape displays, verified the adjustability of axial and lateral stiffness in a 5x5 unit array, and demonstrated the application in an inflatable LDPE capsule.
Embedded Active Stiffening Mechanisms to Modulate Kresling Tower Kinetostatic Properties
J. Berre, P. Renaud
Physics Related
🎯 What it does: Embedding an actuator with adjustable stiffness of the folded face in the Kresling tower to achieve active regulation of axial stiffness and steady-state switching force.
Embodied Agents for Efficient Exploration and Smart Scene Description
Roberto Bigazzi, R. Cucchiara
GenerationRobotic IntelligenceAgentic AIVision-Language-Action ModelImageText
🎯 What it does: This study addresses visual navigation scenarios by constructing an autonomous agent capable of exploring, mapping, and generating natural language descriptions of interesting scenes observed in unknown indoor environments.
Embodied Referring Expression for Manipulation Question Answering in Interactive Environment
Qie Sima, Huaping Liu
Robotic IntelligenceVision-Language-Action ModelWorld ModelMultimodalityBenchmark
🎯 What it does: Proposed the REMQA task and constructed a benchmark dataset, proposing and evaluating a framework that combines 3D semantic reconstruction with modular networks.
EMS®: A Massive Computational Experiment Management System towards Data-driven Robotics
Qinjie Lin, Han Liu
Computational EfficiencyRobotic Intelligence
🎯 What it does: Propose EMS®, a cloud-supported large-scale computational experiment management system designed for high-throughput computing in robotics research.
Emulating Human Kinematic Behavior on Lower-Limb Prostheses via Multi-Contact Models and Force-Based Nonlinear Control
Rachel Gehlhar, A. Ames
OptimizationRobotic Intelligence
🎯 What it does: Achieve human-like kinematic behavior in active transfemoral prosthetic knees, including ankle push-off, using a multi-contact model and force-based nonlinear optimal controller.
Enable Natural Tactile Interaction for Robot Dog based on Large-format Distributed Flexible Pressure Sensors
Lishuang Zhan, Jiangtao Gong
ClassificationRobotic IntelligenceConvolutional Neural NetworkTransformer
🎯 What it does: Designed and implemented a large-scale distributed flexible pressure sensor on a robot dog to achieve natural human-robot tactile interaction.
Enabling safe walking rehabilitation on the exoskeleton Atalante: experimental results
M. Brunet, N. Petit
OptimizationRobotic Intelligence
🎯 What it does: Propose a control architecture that enables gait-impaired patients to actively participate in walking while relaxing constraints on the swing leg's degrees of freedom.
Energy-Based Models for Cross-Modal Localization using Convolutional Transformers
Alan Wu, M. Ryoo
Autonomous DrivingConvolutional Neural NetworkTransformerImagePoint Cloud
🎯 What it does: Propose a cross-modal localization framework based on energy-based models (EBMs), utilizing LiDAR and satellite images to achieve ground vehicle positioning in GPS-denied environments.
Enforcing Constraints for Dynamic Obstacle Avoidance by Compliant Robots
Leonidas Koutras, G. Rovithakis
OptimizationSafty and PrivacyRobotic Intelligence
🎯 What it does: Propose a control scheme that enables the entire body of an active compliance robot to perform dynamic obstacle avoidance constraints, ensuring compliance and accuracy throughout the task process.
Enforcing safety for vision-based controllers via Control Barrier Functions and Neural Radiance Fields
Mukun Tong, Chuchu Fan
Autonomous DrivingSafty and PrivacyNeural Radiance Field
🎯 What it does: Integrate Control Barrier Functions (CBF) with Neural Radiance Fields (NeRF) to ensure the safety of a visual feedback controller by filtering out unsafe actions and intervening to maintain safety.
Enforcing the consensus between Trajectory Optimization and Policy Learning for precise robot control
Quentin Le Lidec, Justin Carpentier
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes an improved method combining reinforcement learning (RL) with trajectory optimization (TO), utilizing sensitivity information generated by TO for Sobolev learning, and employing augmented Lagrangian (AL) techniques to enforce consistency between TO and policy learning, thereby accelerating the learning of global control strategies.
Enhanced Balance for Legged Robots Using Reaction Wheels
Chia-Yen Lee, Zachary Manchester
OptimizationRobotic Intelligence
🎯 What it does: Proposed and implemented a reaction wheel system to enhance the balance and stability of quadruped robots in challenging locomotion tasks.
Enhancing the Efficacy of Lower-body Assistive Devices Through the Understanding of Human Movement in the Real World
Loubna Baroudi, Kira Barton
OptimizationTabularTime SeriesBiomedical Data
🎯 What it does: Combining weekly-scale free walking measurements with laboratory data to quantitatively assess the steady-state proportion of gait and whether speed selection in real-life scenarios aligns with minimum energy expenditure
Ensembles of Compact, Region-specific & Regularized Spiking Neural Networks for Scalable Place Recognition
Somayeh Hussaini, Tobias Fischer
RecognitionAutonomous DrivingSpiking Neural NetworkImageBenchmark
🎯 What it does: Proposed a dense, region-specific spiking neural network (SNN) architecture based on modular integration, where each small local network is responsible for identifying locations within its assigned region, and overactive neurons are removed through regularization methods during the environment learning phase to improve accuracy.
Environment Optimization for Multi-Agent Navigation
Zhan Gao, Amanda Prorok
OptimizationReinforcement Learning
🎯 What it does: Propose a multi-agent navigation environment optimization problem where the environment is treated as a decision variable, and achieve optimization of online/offline and discrete/continuous environment representations through model-agnostic reinforcement learning, ensuring the completeness of navigation goals for all agents.
Epistemic Prediction and Planning with Implicit Coordination for Multi-Robot Teams in Communication Restricted Environments
Lauren Bramblett, N. Bezzo
Robotic Intelligence
🎯 What it does: Propose a coordinated cognitive prediction and planning framework for multi-robot teams in communication-limited environments, enabling consensus without communication and completing applications such as exploration, coverage, task discovery and completion, and convergence.
ERASE-Net: Efficient Segmentation Networks for Automotive Radar Signals
Shihong Fang, Ryan Wu
SegmentationAutonomous DrivingComputational EfficiencyRepresentation LearningPoint Cloud
🎯 What it does: Proposed an efficient automotive radar signal semantic segmentation network called ERASE-Net, which employs a detect-then-segment method to first detect object center points and then extract compact radar signal representations for segmentation.
Error-Domain Conservativity Control to Transparently Increase the Stability Range of Time-Discretized Controllers
Michael Rothammer, J. Ryu
🎯 What it does: By recording the energy value when the error first occurs, generating an energy lower bound through linear interpolation, and enforcing this lower bound using adaptive damping, instability caused by energy decreasing over time under time-discretized control is prevented.
Estimating Tactile Models of Heterogeneous Deformable Objects in Real Time
Shaoxiong Yao, Kris K. Hauser
Computational EfficiencyRobotic Intelligence
🎯 What it does: This paper proposes a method to directly learn the force response of heterogeneous deformable objects from robot sensor data, estimating the force response through volume stiffness field representation and point-based contact simulator.
Estimating the Motion of Drawers From Sound
Manuel Baum, O. Brock
MultimodalityAudio
🎯 What it does: Fusing visual and audio cues to estimate drawer motion
Estimation of continuous environments by robot swarms: Correlated networks and decision-making
Mohsen Raoufi, Heiko Hamann
Robotic Intelligence
🎯 What it does: This paper designs a decentralized robot swarm whose task is to explore an infinite environment, reach consensus on the average of measurable environmental features, and aggregate in regions where this value is measured. The approach involves proposing a control algorithm and validating it through real-world robot swarm experiments in different environments.
Ethical Assessment of a Hospital Disinfection Robot
C. McGinn, P. Treusch
🎯 What it does: Combined the Ethical Canvas (EC) from British Standard 8611 with Ethical Risk Assessment (ERA) methods to evaluate the ethical and socio-cultural impacts of introducing disinfection robots in radiology departments of European hospitals, and formulated mitigation measures
Evaluating Immersive Teleoperation Interfaces: Coordinating Robot Radiation Monitoring Tasks in Nuclear Facilities
H. Stedman, V. Pawar
Robotic IntelligenceSimultaneous Localization and MappingVideo
🎯 What it does: Proposes a VR teleoperation interface for ground robots, featuring dense 3D environment reconstruction and low-latency video streams, and conducts user experiments at the UK Atomic Energy Authority's RACE facility to compare VR and traditional teleoperation interfaces in nuclear facility monitoring and decommissioning tasks.
Evaluating the Feasibility of Magnetic Tools for the Minimum Dynamic Requirements of Microneurosurgery
C. Forbrigger, E. Diller
Robotic IntelligencePhysics Related
🎯 What it does: Designed and implemented a prototype of a serially connected micro-surgical robot driven by magnetic force, and verified its mechanical and dynamic performance through experiments;
Evaluation of Legged Robot Landing Capability Under Aggressive Linear and Angular Velocities
Keran Ye, Konstantinos Karydis
OptimizationRobotic Intelligence
🎯 What it does: Proposes a method to assess the impact landing capability of legged robots under significant contact line velocity and angular velocity.
Event-based Agile Object Catching with a Quadrupedal Robot
Benedek Forrai, D. Scaramuzza
Computational EfficiencyRobotic Intelligence
🎯 What it does: Achieving high-speed ball capturing on a quadruped robot using an event camera, demonstrating the ability to capture objects moving at up to 15 m/s from a distance of 4 meters with an 83% success rate
Event-based Real-time Moving Object Detection Based On IMU Ego-motion Compensation
Chunhui Zhao, Yang Lyu
Object DetectionOptical FlowMultimodality
🎯 What it does: Proposed a real-time moving object detection pipeline based on event cameras and IMU, compensating for camera motion through a nonlinear distortion function, and achieving motion object segmentation via dynamic threshold segmentation, optical flow calculation, and clustering.
Event-Triggered Optimal Formation Tracking Control Using Reinforcement Learning for Large-Scale UAV Systems
Ziwei Yan, Jinjie Li
OptimizationReinforcement Learning
🎯 What it does: Proposed an event-triggered optimal formation tracking controller and constructed a mixed reality experimental platform to verify its effectiveness.
EWareNet: Emotion-Aware Pedestrian Intent Prediction and Adaptive Spatial Profile Fusion for Social Robot Navigation
V. Narayanan, Aniket Bera
Robotic IntelligenceTransformerReinforcement LearningMultimodality
🎯 What it does: Proposes an emotion-aware pedestrian intent prediction and adaptive spatial profile fusion algorithm for social robot navigation.
Ex(plainable) Machina: how social-implicit XAI affects complex human-robot teaming tasks
Marco Matarese, A. Sciutti
Explainability and InterpretabilityRobotic Intelligence
🎯 What it does: In the context of social human-robot interaction, the study uses the Connect 4 game as a complex decision-making task to investigate the impact of counterfactual explanations based on shared experiences on participants' performance and robot persuasiveness; it compares two explanation generation strategies—classical explanations and shared experience-based explanations—to evaluate team performance, robot persuasiveness, and participants' perceptions of the robot and themselves; it also observes that low-performing participants are more inclined to follow the robot.
ExAug: Robot-Conditioned Navigation Policies via Geometric Experience Augmentation
N. Hirose, S. Levine
Data SynthesisRobotic IntelligenceReinforcement LearningPoint Cloud
🎯 What it does: Propose the ExAug framework, which extracts 3D information in point cloud form to perform empirical augmentation for multi-source data from different robot platforms. It achieves cross-platform control strategy training through synthetic image generation and geometric-aware penalties; the trained strategies are evaluated on two new robot platforms and three camera configurations in indoor and outdoor obstacle environments.
EXOT: Exit-aware Object Tracker for Safe Robotic Manipulation of Moving Object
Hyunseo Kim, Byoung-Tak Zhang
Object TrackingAnomaly DetectionRobotic IntelligenceVideo
🎯 What it does: Developed an exit-perception object tracker EXOT, which uses a robot's hand-mounted camera to detect whether the target object disappears during operation, thereby deciding whether to continue the operation.
Expanding the Deployment Envelope of Behavior Prediction via Adaptive Meta-Learning
B. Ivanovic, M. Pavone
Domain AdaptationAutonomous DrivingMeta Learning
🎯 What it does: Proposed a new method that can efficiently transfer behavior prediction models to new environments.
Expanding Versatility of Agile Locomotion through Policy Transitions Using Latent State Representation
G. Christmann, Wei-Chao Chen
Representation LearningRobotic Intelligence
🎯 What it does: Proposes transition-net, a robust transition strategy that distributes the complexity of different gaits into specialized motion control policies, and unifies these policies into a holistic meta-controller through latent state representations, enabling robots to iteratively expand their skill set and achieve robust transitions between any two policies in the skill library, with a single training process under an hour.
Experimental evaluation of a method for improving experiment design in robot identification
Stefanie A. Zimmermann, M. Norrlöf
OptimizationRobotic Intelligence
🎯 What it does: An improved experimental design method was studied and verified to shorten experimental time in robot identification by selecting optimal robotic arm configurations while maintaining or improving parameter estimation accuracy.
Experimental Validation of Functional Iterative Learning Control on a One-Link Flexible Arm
Sjoerd Drost, C. D. Santina
OptimizationRobotic IntelligencePhysics Related
🎯 What it does: Experimental verification of the feasibility of Functional Iterative Learning Control (FILC) on a single-link flexible arm
Experimental Workflow Implementation for Automatic Detection of Filament Deviation in 3D Robotic Printing Process
Xinrui Yang, R. Merzouki
Anomaly DetectionRobotic Intelligence
🎯 What it does: Proposed and implemented an integrated workflow for automatically detecting filament deviations in the 3DCP process, and validated its feasibility through on-site printing tests.
Experiments in Underwater Feature Tracking with Performance Guarantees Using a Small AUV
Benjamin Biggs, D. Stilwell
OptimizationRobotic Intelligence
🎯 What it does: Conduct experiments using small autonomous underwater vehicles (AUVs) in confined areas to locate isobaths.
Expert-Agnostic Ultrasound Image Quality Assessment using Deep Variational Clustering
Deepak Raina, S. Saha
Representation LearningAuto EncoderImageBiomedical DataUltrasound
🎯 What it does: Proposed an unsupervised ultrasound image quality assessment network called US2QNet, which can jointly enhance, extract, cluster, and visualize quality feature representations of ultrasound images.
Explain What You See: Open-Ended Segmentation and Recognition of Occluded 3D Objects
H. Ayoobi, B. Verheij
RecognitionSegmentation
🎯 What it does: Proposes a novel semantic 3D object part segmentation method, combined with an argument-based online incremental learning approach to enhance robustness for 3D object classification in high-occlusion scenarios.
Explainable Action Advising for Multi-Agent Reinforcement Learning
Yue (Sophie) Guo, K. Sycara
Explainability and InterpretabilityReinforcement Learning
🎯 What it does: Proposed and validated an Explainable Action Advising framework, enabling teachers to provide explanations alongside action recommendations, helping students generalize better and improve sample efficiency and learning performance in both single and multi-agent reinforcement learning environments.
Explainable Action Prediction through Self-Supervision on Scene Graphs
Pawit Kochakarn, L. Kunze
Autonomous DrivingExplainability and InterpretabilityTransformerContrastive LearningGraph
🎯 What it does: Explore scene graphs as compressed representations of high-level information for predicting future driver actions, and propose a self-supervised pipeline to obtain representative and disentangled embeddings.
Exploiting Intrinsic Kinematic Null Space for Supernumerary Robotic Limbs Control
T. L. Baldi (University of Siena), D. Prattichizzo (University of Siena)
Robotic Intelligence
🎯 What it does: Leverage the intrinsic kinematic null space of human motion chains to control the additional degrees of freedom of redundant mechanical fingers, achieving better synergy with natural fingers.
Exploiting Trust for Resilient Hypothesis Testing with Malicious Robots
Matthew Cavorsi, Stephanie Gil
Robotic Intelligence
🎯 What it does: Developed a framework that utilizes random trust observations between robots to achieve recoverable binary hypothesis testing in multi-robot crowdsourcing tasks where malicious robots may be present in the majority, and proposed two algorithms.
Exploring An External Approach to Subretinal Drug Delivery via Robot Assistance and B-Mode OCT
Elan Z. Ahronovich, N. Simaan
Robotic IntelligenceDrug DiscoveryBiomedical Data
🎯 What it does: Propose a system that achieves subretinal injection through the assistance of a robot, utilizing an external path to inject medication into the space between the retina and choroid.
Exploring Multimodal Gait Rehabilitation and Assistance through an Adaptable Robotic Platform
S. Otálora, C. Cifuentes
Robotic IntelligenceMultimodalityBiomedical Data
🎯 What it does: Developed and evaluated the adaptable multimodal gait rehabilitation and assistive platform AGoRA V2, which integrates a smart walker and a unilateral lower-limb exoskeleton. The platform's performance was physiologically, kinematically, and perceptually assessed in 14 healthy subjects under four experimental conditions (no device, exoskeleton+walker, walker only, full platform).
Exploring Navigation Maps for Learning-Based Motion Prediction
Julian Schmidt, K. Dietmayer
Autonomous DrivingKnowledge Distillation
🎯 What it does: Explore the use of navigation maps as an alternative to high-precision maps for learning-based motion prediction, and propose a method to integrate navigation maps into prediction models.
Exploring Robot-Assisted Optical Coherence Elastography for Surgical Palpation
Yeonhee Chang, Cheol Song
Robotic IntelligenceBiomedical Data
🎯 What it does: Explores robot-assisted optical coherence elastography (OCE) for surgical palpation, demonstrating an OCE system integrated on a Cartesian Stepping Robot for local stiffness mapping and tumor boundary identification.
External Camera-Based Mobile Robot Pose Estimation for Collaborative Perception with Smart Edge Sensors
S. Bultmann, Sven Behnke
Pose EstimationRobotic IntelligenceConvolutional Neural NetworkSimultaneous Localization and MappingImage
🎯 What it does: Online detection of robot key points through a multi-view RGB camera network and intelligent edge sensors, followed by estimating the pose of mobile robots in a global coordinate system via reprojection error minimization at a central backend, and subsequently fusing their observations into a global scene model;
External Force Estimation of Legged Robots via a Factor Graph Framework with a Disturbance Observer
J. Kang, Kyung-soo Kim
Robotic Intelligence
🎯 What it does: Propose an external force estimation method without using force sensors, obtaining the force of each leg through a disturbance observer, and utilizing preintegration technology to tightly couple external force estimation with pose estimation within a factor graph framework, achieving accurate external force estimation during standing and walking.
Extracting generalizable skills from a single plan execution using abstraction-critical state detection
Khen Elimelech, M. Vardi
Optimization
🎯 What it does: Proposed a method for automatically extracting reusable and generalizable abstract skills that can learn from a single plan execution and be reused in new domains
Extraneousness-Aware Imitation Learning
Rachel Zheng, Huazhe Xu
RetrievalReinforcement Learning from Human FeedbackVideo
🎯 What it does: Proposed a self-supervised visual imitation learning method called EIL for learning visual motion policies from third-person demonstrations that contain additional irrelevant subsequences.
Extremum Seeking-Based Adaptive Sliding Mode Control with Sliding Perturbation Observer for Robot Manipulators
H. Khan, Minwoo Lee
OptimizationRobotic Intelligence
🎯 What it does: Proposed an adaptive robust sliding mode control (SMC) combined with a nonlinear sliding mode disturbance observer (SPO) for robotic manipulator control, utilizing extremal optimization (ES) to achieve real-time adaptive control gain adjustment
Extrinsic calibration for highly accurate trajectories reconstruction
Maxime Vaidis, F. Pomerleau
Pose EstimationSequential
🎯 What it does: A novel external calibration method for multiple robotic total stations is proposed, which does not require manual ground control points or synchronized measurements, to accurately obtain the pose of three-degree-of-freedom targets in a six-degree-of-freedom coordinate system in field environments.
Factor Graph Fusion of Raw GNSS Sensing with IMU and Lidar for Precise Robot Localization without a Base Station
J. Beuchert, M. Fallon
Autonomous DrivingOptimizationRobotic IntelligenceSimultaneous Localization and MappingMultimodalityPoint CloudTime Series
🎯 What it does: Propose a tightly coupled factor graph framework that fuses raw GNSS data, inertial measurements, and optional LiDAR to achieve precise and smooth mobile robot localization.