ICRA 2023 Papers — Page 13
IEEE International Conference on Robotics and Automation · 1341 papers
Task-Directed Exploration in Continuous POMDPs for Robotic Manipulation of Articulated Objects
Aidan Curtis, Siddarth Jain
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes STRUG, an online POMDP solver capable of handling both task-related and task-irrelevant uncertainties and supporting long-term planning, validated on extended toy POMDP problems and robot joint object manipulation tasks using a neural perception frontend to construct model distributions.
Task-Driven Graph Attention for Hierarchical Relational Object Navigation
Michael Lingelbach, Jiajun Wu
Autonomous DrivingGraph Neural NetworkGraph
🎯 What it does: Propose a task-driven attention mechanism that combines scene graphs with graph neural networks (GNNs) to address hierarchical relationship object navigation tasks
Task-Oriented Stiffness Setting for a Variable Stiffness Hand
Ana Elvira H. Martin, M. Roa
OptimizationRobotic Intelligence
🎯 What it does: Proposed a method to set optimal stiffness for a multi-fingered variable stiffness hand to accomplish unknown force tasks in a specified direction; analyzed force distribution among fingers using endpoint stiffness ellipsoid analysis, and verified its effectiveness through an iterative adaptive strategy in a door opening experiment.
Task-Space Clustering for Mobile Manipulator Task Sequencing
Quang-Nam Nguyen, Quang Pham
OptimizationRobotic Intelligence
🎯 What it does: Proposed a task space clustering method that models the clustering process as a set cover problem utilizing bidirectional graphs and reachability analysis to achieve the minimum number of target clusters and corresponding base placement schemes; applied to hundreds of targets in mobile drilling experiments; demonstrated superior solutions and faster computation time through multiple simulations compared to existing state-of-the-art methods.
Telerobot operators can account for varying transmission dynamics in a visuo-haptic object tracking task
Mohit Singhala, Jeremy D. Brown
Robotic IntelligenceMultimodality
🎯 What it does: Using 12 participants to perform visual-tactile tracking tasks on a reconfigurable teleoperator with variable transmission dynamics to investigate how humans compensate for these dynamics.
Temporal Logic Swarm Control with Splitting and Merging
G. Cardona, C. Vasile
OptimizationRobotic Intelligence
🎯 What it does: Designed a communication-free multi-robot swarm control framework based on Metric Temporal Logic (MTL), which dynamically adjusts subgroup numbers through splitting and merging during tasks, and realizes planning and allocation of subgroup trajectories and splitting ratios via a two-phase process.
Tendon-Driven Soft Robotic Gripper with Integrated Ripeness Sensing for Blackberry Harvesting
Alex Qiu, Ai-Ping Hu
Robotic IntelligenceImageAgriculture Related
🎯 What it does: Designed and manufactured a tendon-driven flexible grasping system for automatic blackberry picking, integrating near-infrared (NIR) fruit ripeness sensors, visual servo cameras, and micro-servos for tendon contraction;
Tentacle-Based Shape Shifting of Metamorphic Robots Using Fast Inverse Kinematics
J. Mrázek, J. Barnat
Robotic Intelligence
🎯 What it does: Propose a deformation-based reconfiguration method for a deformable robot, first transforming a chain-type robot into an octopus-like multi-tentacle shape, then gradually reconnecting the tentacles through inverse kinematics to form a snake-like structure, and finally inversely achieving the target morphology reconfiguration.
Test-time Domain Adaptation for Monocular Depth Estimation
Zhi Li, Dengxin Dai
Depth EstimationDomain AdaptationSupervised Fine-TuningImage
🎯 What it does: Proposes a test-time domain adaptation framework for monocular depth estimation that can instantly adapt the source pre-trained model to test data in a source-free and unsupervised manner.
Test-Time Synthetic-to-Real Adaptive Depth Estimation
Eojindl Yi, Junmo Kim
Depth EstimationDomain AdaptationAutonomous DrivingImage
🎯 What it does: Proposed a single-image depth estimation method that can continuously adapt to domain drift during testing
Testing Rare Downstream Safety Violations via Upstream Adaptive Sampling of Perception Error Models
Craig Innes, S. Ramamoorthy
Autonomous DrivingSafty and PrivacyImage
🎯 What it does: This paper combines the perception error model with state-dependent adaptive importance sampling to efficiently evaluate the rare failure probability of real perception control systems in simulation environments.
The Human Gaze Helps Robots Run Bravely and Efficiently in Crowds
Qianyi Zhang, Jingtai Liu
Robotic IntelligenceImage
🎯 What it does: Propose an improved limit cycle for cooperatively parameterizing human intent and planning robot motion, incorporating human gaze information into a dynamic chicken game model;
The New Dexterity Adaptive Humanlike Robot Hand: Employing a Reconfigurable Palm for Robust Grasping and Dexterous Manipulation
Geng Gao, Minas Liarokapis
Robotic Intelligence
🎯 What it does: Design and verify two humanoid robotic hands capable of executing robust encircling grasps under various environmental uncertainties, enhancing grasping and manipulation capabilities through reconfigurable palm and thumb base degrees of freedom.
The new exhibition Blind machines, a large 3D printing machine
J. Merlet, Y. Papegay
Robotic Intelligence
🎯 What it does: Developed a large-scale 3D printer based on a 3-degree-of-freedom cable-driven parallel robot (CDPR) for art exhibitions.
The Reflectance Field Map: Mapping Glass and Specular Surfaces in Dynamic Environments
P. Foster, B. Kuipers
Computational EfficiencySimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed a LiDAR-based reflectance field map that can real-time detect mirror-reflective surfaces such as glass, metal, and mirrors, integrating light field mapping theory with occupancy grid mapping.
The Role of Symmetry in Constructing Geometric Flat Outputs for Free-Flying Robotic Systems
Jake Welde, Vijay R. Kumar
Robotic Intelligence
🎯 What it does: Construct the geometric flat output of the free-flying robot system based on symmetry, and provide sufficient conditions and examples
The SLAM Hive Benchmarking Suite
Yuanyuan Yang, Sören Schwertfeger
Simultaneous Localization and MappingBenchmark
🎯 What it does: Proposes a benchmark testing framework named SLAM Hive, which utilizes container technology and cluster deployment to systematically analyze SLAM algorithms across thousands of mapping runs.
The Sum of Its Parts: Visual Part Segmentation for Inertial Parameter Identification of Manipulated Objects
Philippe Nadeau, Jonathan Kelly
SegmentationPoint CloudMesh
🎯 What it does: Propose an algorithm that identifies the inertial parameters of manipulated objects by combining visual segmentation with torque sensing, requiring only slow or paused motion;
The Third Generation (G3) Dual-Modal and Dual Sensing Mechanisms (DMDSM) Pretouch Sensor for Robotic Grasping
Cheng Fang, Jun Zou
Robotic IntelligenceMultimodalityUltrasound
🎯 What it does: Developed and verified a third-generation dual-mode dual-sensing pre-touch sensor (G3 DMDSM) based on pulse-echo ultrasound and optoacoustic imaging for close-range ranging and material perception in robotic grasping.
Throwing Objects into A Moving Basket While Avoiding Obstacles
H. Kasaei, M. Kasaei
Robotic IntelligenceReinforcement Learning
🎯 What it does: Studied the technique of robots throwing objects into a moving basket in the presence of obstacles.
TJ-FlyingFish: Design and Implementation of an Aerial-Aquatic Quadrotor with Tiltable Propulsion Units
Xuchen Liu, Ben M. Chen
Robotic Intelligence
🎯 What it does: Designed and implemented an aerial-underwater quadrotor prototype equipped with a switchable dual-speed propulsion system and a rotatable propulsion unit to achieve dual-mode kinetic and thrust vector control in underwater and aerial environments.
TODE-Trans: Transparent Object Depth Estimation with Transformer
Kan Chen, Bin Li
Depth EstimationTransformerImage
🎯 What it does: Developed a Transformer-based method for depth estimation of transparent objects, using a single RGB-D input to predict the surface depth of transparent objects
TOFG: A Unified and Fine-Grained Environment Representation in Autonomous Driving
Zihao Wen, Jianping Wang
Autonomous DrivingGraph Neural NetworkGraphSequential
🎯 What it does: Proposed and implemented a unified and fine-grained environmental representation method called Temporal Occupancy Flow Graph (TOFG), combined with Graph Attention Network (GAT) for trajectory prediction and motion planning.
TOP-JAM: A bio-inspired topology-based model of joint attention for human-robot interaction
H. F. Chame, R. Alami
Explainability and InterpretabilityRobotic Intelligence
🎯 What it does: Proposed a topology-based joint attention model called TOP-JAM for real-time tracking of joint attention states in human-robot interaction
Topological Trajectory Prediction with Homotopy Classes
Jennifer Wakulicz, R. Fitch
Object TrackingRepresentation LearningTime SeriesSequential
🎯 What it does: Propose a lightweight learning framework that utilizes a variable-order Markov process to predict the homotopy classes of trajectories, enabling high-level trajectory prediction, and integrates homotopy class information into a Gaussian Mixture Model (GMM) to enhance low-level trajectory prediction;
Topology Matching of Branched Deformable Linear Objects
Manuel Zürn, A. Verl
Pose EstimationGraph Neural NetworkImageGraph
🎯 What it does: Propose a graph-based segment matching method to estimate correspondence between known branching variable linear object topology and image representations captured by 3D stereo cameras.
Topology-Based MPC for Automatic Footstep Placement and Contact Surface Selection
Jae-Koo Shim, S. Vijayakumar
OptimizationRobotic Intelligence
🎯 What it does: Proposed a topology-based MPC framework capable of real-time simultaneous planning of full-body motion, torque commands, foot gaits, and contact surface selection.
Torque Control with Joints Position and Velocity Limits Avoidance
Venus Pasandi, D. Pucci
Robotic Intelligence
🎯 What it does: Designed a torque control architecture to enable a fully driven robotic arm to track desired time-varying trajectories while ensuring joint position and velocity limits.
Torque-Limited Manipulation Planning through Contact by Interleaving Graph Search and Trajectory Optimization
Ramkumar Natarajan, Howie Choset
OptimizationRobotic Intelligence
🎯 What it does: A whole-arm motion planning method that utilizes environmental contact to expand the reachable workspace of torque-limited robotic arms was studied
Touch Classification on Robotic Skin using Multimodal Tactile Sensing Modules
Minjin Yang, Jung Kim
ClassificationConvolutional Neural NetworkMultimodality
🎯 What it does: Developed a robotic skin equipped with a multimodal tactile sensing module, achieving extensive spatiotemporal perception capabilities using a minimal number of sensing elements.
Toward Cooperative 3D Object Reconstruction with Multi-agent
Xiong Li, Zhen Hong
GenerationPose EstimationNeural Radiance FieldAuto EncoderImage
🎯 What it does: Propose a multi-agent collaborative 3D object reconstruction method, which uses multiple visual sensors to decompose a complete object into local 3D models, restores them by different agents, and then integrates the local models by estimating relative poses.
Toward Efficient Physical and Algorithmic Design of Automated Garages
Teng Guo, Jingjin Yu
OptimizationPhysics Related
🎯 What it does: Propose an automated garage design that nearly achieves 100% parking density, modeling the multi-vehicle parking and retrieval process as a multi-robot path planning problem. Optimal and approximate algorithms, along with a novel shuffling mechanism, are designed to enable scheduled retrieval during peak hours.
Toward Fine Contact Interactions: Learning to Control Normal Contact Force with Limited Information
Jinda Cui, J. Trinkle
Robotic IntelligenceReinforcement Learning
🎯 What it does: A normal contact force controller was trained using model-free reinforcement learning on low-cost, information-poor tactile sensors, and combined with a motion controller to achieve object manipulation in non-grasping, fine contact interactions.
Toward Zero-Shot Sim-to-Real Transfer Learning for Pneumatic Soft Robot 3D Proprioceptive Sensing
Uksang Yoo, Chen Feng
Domain AdaptationRobotic IntelligenceImagePoint Cloud
🎯 What it does: Proposed and verified a robust sim-to-real transmission pipeline for collecting full-body shape information of soft robots under high-fidelity point cloud representations, and evaluated the model directly on real internal camera images after training on simulated data.
Towards a Finned-Swimming Exoskeleton: A Robotic Flutter Kicking Testbed and its Corresponding Thrust Generation
Beau Johnson, M. Goldfarb
Robotic Intelligence
🎯 What it does: Designed a robot platform for fin-based swimming and developed a controller to generate fin-based swimming motions; conducted experiments and recorded thrust generation results produced by flutter kicks.
Towards a Reliable and Lightweight Onboard Fault Detection in Autonomous Unmanned Aerial Vehicles
Sai Srinadhu Katta, E. Viegas
Anomaly DetectionComputational Efficiency
🎯 What it does: Propose a lightweight drone physical fault detection model based on machine learning, which can detect faults with high precision and low computational cost, and promptly notify operators when the model fails.
Towards Autonomous UAV Railway DC Line Recharging: Design and Simulation
Frederik Falk Nyboe, E. Ebeid
Robotic Intelligence
🎯 What it does: Proposed a drone system capable of autonomously charging from railway DC lines, equipped with dual power line clamps and a cable drum to achieve two-stage landing and energy capture;
Towards Bridging the Space Domain Gap for Satellite Pose Estimation using Event Sensing
A. M. Jawaid, Tat-Jun Chin
Data SynthesisPose Estimation
🎯 What it does: Train a domain-agnostic close-range satellite pose estimation model using synthetic event data and evaluate it on a laboratory real event dataset.
Towards Consistent Batch State Estimation Using a Time-Correlated Measurement Noise Model
David J. Yoon, T. Barfoot
OptimizationRobotic IntelligenceSupervised Fine-TuningTime SeriesSequential
🎯 What it does: This paper proposes an algorithm for batch state estimation by learning time-related measurement covariance matrices, and parameterizes the inverse measurement covariance matrix as a block-banded structure to improve computational efficiency.
Towards Efficient Gas Leak Detection in Built Environments: Data-Driven Plume Modeling for Gas Sensing Robots
Wanting Jin, A. Martinoli
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Proposes a new method that combines a learning-based plume model with existing probabilistic GSL algorithm Source Term Estimation (STE) to achieve efficient localization of gas sources in building environments.
Towards Efficient Trajectory Generation for Ground Robots beyond 2D Environment
Jingping Wang, Fei Gao
OptimizationRobotic Intelligence
🎯 What it does: Proposed an optimization-based trajectory planning framework that can simultaneously handle active and passive height changes of ground robots in three-dimensional space.
Towards Exact Interaction Force Control for Underactuated Quadrupedal Systems with Orthogonal Projection and Quadratic Programming
Shengzhi Wang, K. W. S. Au
OptimizationRobotic Intelligence
🎯 What it does: A new scheme for interaction force control in underactuated quadruped robots is designed, utilizing projection technology and quadratic programming to achieve precise interaction force output without the need for force sensors, while minimizing base motion.
Towards Generalized Robot Assembly through Compliance-Enabled Contact Formations
A. S. Morgan, A. Dollar
Robotic IntelligenceTime Series
🎯 What it does: A compatibility-enabled contact formation method is used to enable robots to control constraints in insertion tasks without explicitly estimating contact positions, instead controlling constraints by monitoring the force of the end-effector.
Towards Human-Robot Collaboration with Parallel Robots by Kinetostatic Analysis, Impedance Control and Contact Detection
Aran Mohammad, T. Ortmaier
Robotic Intelligence
🎯 What it does: Studied and verified collision detection and impedance control methods for parallel robots when collaborating with humans, employing a generalized momentum-based disturbance observer, Kalman filter, and second-order sliding mode observer to estimate external forces and achieve low-impedance control;
Towards Multi-Day Field Deployment Autonomy: A Long-Term Self-Sustainable Micro Aerial Vehicle Robot
S. Carlson, C. Papachristos
Robotic Intelligence
🎯 What it does: Proposed a self-sustaining micro aerial vehicle system architecture for multi-day field deployment, achieving functionalities such as deep sleep, continuous operation, multi-day endurance, and external power charging, with a zero-intervention multi-day field deployment demonstration conducted in Nevada.
Towards Open-Set Material Recognition using Robot Tactile Sensing
Kunhong Liu, Xiangyi Huang
ClassificationRobotic IntelligenceConvolutional Neural NetworkGenerative Adversarial Network
🎯 What it does: Propose an open-set material recognition framework based on GCPL and GAN, validated on two batches of tactile data collected by electronic skin.
Towards Open-World Interactive Disambiguation for Robotic Grasping
Yuchen Mo, Hanbo Zhang
Robotic IntelligenceTransformerLarge Language ModelVision Language ModelImageText
🎯 What it does: Proposes the SeeAsk system, which performs interactive visual localization for target grasping in open-world environments using ambiguous natural language instructions.
Towards Predicting Fine Finger Motions from Ultrasound Images via Kinematic Representation
Dean Zadok, A. Bronstein
Pose EstimationRepresentation LearningVideoBiomedical DataUltrasound
🎯 What it does: Developed an end-to-end system that utilizes ultrasound image sequences and hand kinematic representations to infer fine-grained fingertip movements
Towards Robots that Influence Humans over Long-Term Interaction
Shahabedin Sagheb, Dylan P. Losey
Robotic Intelligence
🎯 What it does: This paper studies the impact of robots on humans during long-term interactions, pointing out that humans adapt to robots and predict their behaviors, thereby reducing the influence of traditional Stackelberg games. Subsequently, three improved Stackelberg game methods are proposed to make robot strategies more influential and harder to predict. These improvements are validated through simulations and user experiments, demonstrating their effectiveness in maintaining long-term influence.
Towards Robust Autonomous Grasping with Reflexes Using High-Bandwidth Sensing and Actuation
Andrew SaLoutos, Sangbae Kim
Robotic Intelligence
🎯 What it does: Developed an autonomous grasping reflex control system based on high-frequency force, contact, and distance perception, combined with low-frequency trajectory planning, achieving autonomous grasping for high-speed, low-inertia robotic arms.
Towards Robust Reference System for Autonomous Driving: Rethinking 3D MOT
Leichen Wang, Xinrun Lil
Object TrackingAutonomous DrivingPoint Cloud
🎯 What it does: In this paper, we propose a set of innovative 3D multi-object tracking post-processing modules to construct a unified framework, mainly including a self-learning detector, the GGTrajRec module for recovering trajectory breakpoints and ID switches, and a confidence-based trajectory optimizer.
Towards Safe Landing of Falling Quadruped Robots Using a 3-DoF Morphable Inertial Tail
Yunxi Tang, K. W. S. Au
Robotic Intelligence
🎯 What it does: Proposed a 3-degree-of-freedom deformable inertial tail for controlling a falling quadruped robot during flight, enabling self-righting in mid-air and achieving safe landing.
Towards Safe Remote Manipulation: User Command Adjustment based on Risk Prediction for Dynamic Obstacles
Mincheul Kang, Sung-eui Yoon
Robotic Intelligence
🎯 What it does: Propose a risk-aware user command adjustment method that utilizes a network to predict dynamic obstacle risks and synthesize collision-avoidance instructions, making real-time decisions on whether to adjust user commands to ensure safety.
Towards Surgical Context Inference and Translation to Gestures
Kay Hutchinson, H. Alemzadeh
SegmentationRecurrent Neural NetworkImageVideoBiomedical Data
🎯 What it does: Propose a method for automatically generating surgical gesture transcription using image segmentation. First, detect surgical context through segmentation masks, then convert context labels into gesture transcription, achieving interpretability.
Towards True Lossless Sparse Communication in Multi-Agent Systems
Seth Karten, K. Sycara
Representation LearningAuto Encoder
🎯 What it does: Proposes the Information Maximization Gated Sparse Multi-Agent Communication (IMGS-MAC) method, which reframes sparse communication as a representation learning problem under the information bottleneck framework. It achieves lossless sparse communication through two individualized regularization objectives: an information-maximizing autoencoder and a sparse communication loss. In collaborative multi-agent tasks, the learned communication 'language' is directly subjected to causal analysis to determine zero-shot sparse budget ranges and reward loss ranges. A gating function further compresses communication under few-sample conditions. The method also conducts experiments on continuous and discrete messages with multiple ablation studies to validate the model's lossless performance.
Towards Unsupervised Filtering of Millimetre-Wave Radar Returns for Autonomous Vehicle Road Following
Dean Sacoransky, K. Hashtrudi-Zaad
Autonomous DrivingPoint Cloud
🎯 What it does: Utilize millimeter-wave radar to detect roadside reflectors, and apply unsupervised learning DBSCAN to filter noise in radar point clouds, providing vehicle forward road path prediction.
Towards View-invariant and Accurate Loop Detection Based on Scene Graph
Chuhao Liu, S. Shen
Simultaneous Localization and Mapping
🎯 What it does: Proposed an indoor visual SLAM loop detection method based on incremental construction of scene graphs.
Towards Visual Classification Under Class Ambiguity
V. Kozák, L. Přeučil
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: A systematic comparison of multiple CNN classification variants was conducted in class-ambiguous image data interpretation scenarios, introducing prior fuzzy information as soft true value labels to improve classification accuracy, while proposing an interpretable method based on Bayesian CNN and validating it on practical weld inspection tasks.
TRADE: Object Tracking with 3D Trajectory and Ground Depth Estimates for UAVs
Pedro F. Proença, J. Delaune
Object TrackingSegmentationDepth Estimation
🎯 What it does: Propose the TRADE method for robust tracking and 3D localization of moving targets in complex environments using a UAV single-camera setup, supporting 3D-aware target following.
Traffic-Aware Autonomous Driving with Differentiable Traffic Simulation
L. Zheng, Ming-Chyuan Lin
Autonomous DrivingOptimizationKnowledge Distillation
🎯 What it does: Proposes the Traffic-Aware Autonomous Driving (TrAAD) method, which combines traffic simulation information with deep learning-based imitation learning to directly optimize traffic flow speed and energy consumption, focusing on speed control supervision;
TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Prediction
Zhejun Zhang, L. Gool
Autonomous DrivingTransformerWorld Model
🎯 What it does: Propose TrafficBots, a multi-agent strategy based on motion prediction and end-to-end driving, treating data-driven traffic simulation as a world model to achieve configurable and scalable traffic simulation, with experiments conducted on the Waymo Open Motion Dataset.
TrafficGen: Learning to Generate Diverse and Realistic Traffic Scenarios
Lang Feng, Bolei Zhou
Data SynthesisAutonomous DrivingTransformerTime SeriesSequential
🎯 What it does: Developed TrafficGen, a traffic scene generation method based on autoregressive neural generative models, which can learn from fragmented human driving data to generate diverse and realistic traffic scenes, while also adding vehicles and extending trajectories in existing scenes.
Train Offline, Test Online: A Real Robot Learning Benchmark
G. Zhou, Abhi Gupta
Robotic IntelligenceReinforcement LearningVision-Language-Action ModelImageBenchmark
🎯 What it does: Proposes the Train Offline, Test Online (TOTO) benchmark, providing a remotely shared robotic online testing and offline training dataset covering manipulation tasks requiring generalization to unseen objects, positions, and lighting conditions.
Train What You Know – Precise Pick-and-Place with Transporter Networks
G. Sóti, B. Hein
Computational EfficiencyRobotic Intelligence
🎯 What it does: Defined precise training methods and iterative inference methods to improve the accuracy of grasping and placement in Transporter Networks, conducted large-scale experiments on 8 simulated tasks, proposed a network architecture improvement that maintains performance while reducing computational cost and time, and validated the method through interactive teaching on real hardware.
Training Efficient Controllers via Analytic Policy Gradient
Nina Wiedemann, D. Scaramuzza
Computational EfficiencyRobotic IntelligenceReinforcement Learning
🎯 What it does: Propose an Analytical Policy Gradient (APG) method that utilizes a differentiable simulator to train an offline controller for high-precision trajectory tracking;
Trajectory and Sway Prediction Towards Fall Prevention
Weizhuo Wang, Monroe Kennedy
Safty and PrivacyVideoTime Series
🎯 What it does: Propose a metric to monitor the correlation between trunk sway and active/passive perturbations, and predict future paths and trunk sway changes using past trajectories, trunk motion, and surrounding scenes.
Trajectory error compensation for optimal control of UMA-2 – a climbing robot executing maintenance operation in harsh environment
D. Gitardi, A. Valente
OptimizationRobotic Intelligence
🎯 What it does: Proposed and verified a trajectory analysis and adaptive model to control the trajectory of the UMA-2 robot when performing maintenance tasks on vertical and curved surfaces, ensuring surface quality Key Performance Indicators (KPIs).
Trajectory Generation with Dynamic Programming for End-Effector Sway Damping of Forestry Machine
Iman Jebellat, I. Sharf
OptimizationRobotic IntelligenceAgriculture Related
🎯 What it does: Generate anti-swing trajectories using dynamic programming to reduce the swing of the end-effector in forestry machines during fast motion.
Trajectory Optimization for 3D Shape-Changing Robots with Differential Mobile Base
Mengke Zhang, Yanjun Cao
OptimizationRobotic IntelligenceOrdinary Differential Equation
🎯 What it does: Proposes a trajectory optimization method for differential drive mobile robots with controllable deformability in dense 3D environments, modeling full-body trajectories as polynomial curves while satisfying nonholonomic chassis dynamics and additional joint dynamics constraints. Soft constraints and dense sampling activation functions are employed to avoid nonlinear variations while ensuring full-shape safety.
Trajectory Optimization for Distributed Manipulation by Shaping a Physical Field
Adam Uchytil, Jirí Zemánek
OptimizationPhysics Related
🎯 What it does: Propose a distributed manipulation method based on trajectory optimization, utilizing a continuous physical field generated and shaped in real-time by an actuator array to reconfigure multiple independent objects on a plane; first construct the trajectory optimization problem and initialization scheme, then implement this method on a magnetic field distributed manipulation experimental platform, capable of reconfiguring up to eight objects at once while avoiding collisions.
Trajectory Planning for the Bidirectional Quadrotor as a Differentially Flat Hybrid System
Katherine Mao, Vijay R. Kumar
OptimizationRobotic Intelligence
🎯 What it does: A trajectory planning algorithm for bidirectional thruster quadrotors was developed, enabling smooth passage through free-fall singularities and achieving forward and backward thrust switching.
Trajectory planning issues in cuspidal commercial robots
Durgesh Haribhau Salunkhe, P. Wenger
Robotic Intelligence
🎯 What it does: This paper presents a case study on Kinova Robotics' JACO (2nd generation, non-spherical wrist) serial arm, demonstrating the property of cuspidal serial robots where inverse kinematic solutions do not pass through singularities, and provides non-singular solution transformations to highlight the impact of environmental interference on cuspidal robots; it also discusses potential issues and their consequences that may arise when selecting initial solutions in path planning.
Transferring Implicit Knowledge of Non-Visual Object Properties Across Heterogeneous Robot Morphologies
Gyan Tatiya, Jivko Sinapov
ClassificationRecognitionRobotic Intelligence
🎯 What it does: Propose a multi-stage projection framework that can transfer implicit object attribute knowledge between different robot morphologies, evaluate it on object attribute recognition and identity recognition tasks, and introduce data augmentation techniques to enhance model generalization capabilities.
Transparent Objects: A Corner Case in Stereo Matching
Zhiyuan Wu, Rui Fan
Object DetectionSegmentationDepth EstimationAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: Propose a transparent object-aware stereo matching method (TA-Stereo), which first identifies transparent objects using semantic segmentation or salient object detection networks, and then applies homogenization processing to make stereo matching algorithms treat them as non-transparent objects.
TransRSS: Transformer-based Radar Semantic Segmentation
Haojie Zou, Yutao Gao
SegmentationConvolutional Neural NetworkTransformer
🎯 What it does: Proposed a Transformer-based radar semantic segmentation method called TransRSS to effectively and efficiently extract semantic features from radar data.
TransVisDrone: Spatio-Temporal Transformer for Vision-based Drone-to-Drone Detection in Aerial Videos
Tushar Sangam, M. Shah
Object DetectionConvolutional Neural NetworkTransformerVideo
🎯 What it does: Proposes an end-to-end framework named TransVisDrone for drone-to-drone detection in aerial videos.
Tree-structured Policy Planning with Learned Behavior Models
Yuxiao Chen, M. Pavone
Autonomous DrivingReinforcement LearningWorld ModelMultimodalityPoint Cloud
🎯 What it does: Propose Tree Policy Planning (TPP), which transforms continuous optimization problems into solvable discrete Markov Decision Processes (MDPs) by constructing ego trajectory trees and scenario trees, enabling multi-stage motion planning while being compatible with advanced deep learning prediction models.
TTCDist: Fast Distance Estimation From an Active Monocular Camera Using Time-to-Contact
Levi Burner, Y. Aloimonos
Depth EstimationImageMultimodality
🎯 What it does: Proposed and validated two constraints based on temporal contact, acceleration, and distance (τ and Φ) for depth estimation using active monocular cameras and IMUs.
Twist Snake: Plastic table-top cable-driven robotic arm with all motors located at the base link
Kazutoshi Tanaka, Masashi Hamaya
Robotic Intelligence
🎯 What it does: Designed and verified a plastic desktop-level cable-driven robotic arm named Twist Snake, placing all motors in the base, and conducted experimental verification using compact force transmission devices and anti-loosening mechanisms.
Twisting Spine or Rigid Torso: Exploring Quadrupedal Morphology via Trajectory Optimization
J. Caporale, D. Koditschek
OptimizationRobotic Intelligence
🎯 What it does: Compared the energy consumption of one-axis torsional spine and rigid torso in different motions of quadruped robots using trajectory optimization.
Two-Stage Grasping: A New Bin Picking Framework for Small Objects
Hanwen Cao, Yunhui Liu
Object DetectionSegmentationRobotic IntelligenceImage
🎯 What it does: Propose a dual-stage grasping framework for precise grasping of cluttered small objects in box picking.
UDepth: Fast Monocular Depth Estimation for Visually-guided Underwater Robots
Boxiao Yu, M. Islam
Depth EstimationConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposed a fast monocular depth estimation method called UDepth, aiming to provide 3D perception capabilities for low-cost underwater robots.
Ultra-low Power Deep Learning-based Monocular Relative Localization Onboard Nano-quadrotors
Stefano Bonato, D. Palossi
Pose EstimationOptimizationImage
🎯 What it does: This paper designs and implements an end-to-end monocular relative positioning system, utilizing deep neural networks (DNNs) for relative positioning of nano drones with weights below 40g and power consumption below 100mW, completing the entire process from dataset collection, augmentation, quantization to field deployment during actual flight.
Uncertainty Quantification of Collaborative Detection for Self-Driving
Sanbao Su, Fei Miao
Object DetectionAutonomous Driving
🎯 What it does: Proposed and implemented a uncertainty quantification method called Double-M Quantification for collaborative object detection, used to estimate the uncertainty of detection results.
Uncertainty-aware LiDAR Panoptic Segmentation
Kshitij Sirohi, Wolfram Burgard
SegmentationAutonomous DrivingPoint Cloud
🎯 What it does: Propose the EvLPSNet network to address the uncertainty-aware panoptic segmentation problem for LiDAR point clouds, predicting semantic, instance segmentation, and uncertainty estimation for each point.
Uncertainty-Guided Active Reinforcement Learning with Bayesian Neural Networks
Xinyang Wu, Marco F. Huber
Safty and PrivacyReinforcement Learning
🎯 What it does: This paper proposes using Bayesian Neural Networks (BNN) to guide agents in active exploration, aiming to improve the learning efficiency of Reinforcement Learning (RL), and to identify safety risks in the working environment through uncertainty information.
Unidirectional-Road-Network-Based Global Path Planning for Cleaning Robots in Semi-Structured Environments
Yong Li, Hui Cheng
OptimizationRobotic Intelligence
🎯 What it does: Construct a one-way road network to represent traffic constraints in semi-structured environments, propose a hybrid strategy allowing start and end points to cross roads, and employ a two-layer potential field map to handle complex intersections, ensuring planning results.
Unseen Object Instance Segmentation with Fully Test-time RGB-D Embeddings Adaptation
Lu Zhang, Zhiyong Liu
SegmentationDomain AdaptationKnowledge DistillationMultimodality
🎯 What it does: Propose a method for unseen object instance segmentation during testing by adjusting the BatchNorm parameters of a simulated training model.
Unsupervised Learning of Depth and Pose Based on Monocular Camera and Inertial Measurement Unit (IMU)
Yanbo Wang, Yi Huang
Pose EstimationDepth EstimationMultimodality
🎯 What it does: Unsupervised depth and pose estimation using a monocular camera and IMU, with absolute depth and scale factor recovery achieved through IMU acceleration constraints.
Unsupervised Quality Prediction for Improved Single-Frame and Weighted Sequential Visual Place Recognition
Helen Carson, Michael Milford
Retrieval
🎯 What it does: Propose a training-free localization quality prediction method and use the prediction results to weight the sequence matching process to enhance VPR performance.
Unsupervised RGB-to-Thermal Domain Adaptation via Multi-Domain Attention Network
L. Gan, Soon-Jo Chung
ClassificationSegmentationDomain AdaptationRepresentation LearningConvolutional Neural NetworkImage
🎯 What it does: Proposes an unsupervised RGB-to-thermal domain adaptation method that utilizes a multi-domain attention network to achieve thermal image classification and semantic segmentation.
Unsupervised Road Anomaly Detection with Language Anchors
Beiwen Tian, Guyue Zhou
Anomaly DetectionVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: Unsupervised Road Anomaly Detection Using Language Anchors and Scene Parsing Logits
UPLIFT: Unsupervised Person Labeling and Identification via Cooperative Learning with Mobile Robots
Y. Tseng, Fang-jing Wu
RecognitionDomain AdaptationRobotic IntelligenceVideo
🎯 What it does: Propose the UPLIFT framework, which utilizes a mobile robot as a knowledge seed to automatically generate unsupervised person labels and recognition datasets for fixed surveillance cameras.
Upper-limb Geometric MyoPassivity Map for Physical Human-Robot Interaction
Xingyuan Zhou, S. F. Atashzar
Robotic IntelligenceBiomedical Data
🎯 What it does: This study first investigates the passivity mapping of the upper limb through frequency behavior under real-time visual EMG feedback control, and proposes a geometric myopassivity (Geometric MyoPassivity, GMP) mapping model.
USEEK: Unsupervised SE(3)-Equivariant 3D Keypoints for Generalizable Manipulation
Zhengrong Xue, Huazhe Xu
Pose EstimationRepresentation LearningRobotic Intelligence
🎯 What it does: Developed an unsupervised SE(3)-equivariant keypoint method called USEEK for inferring the manipulability of unseen objects of the same class under arbitrary poses from a single demonstrated grasp pose.
User-Conditioned Neural Control Policies for Mobile Robotics
L. Bauersfeld, D. Scaramuzza
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a framework for dynamically modulating vehicle behavior by incorporating auxiliary inputs (such as maximum available thrust or perspective relative to the next waypoint) into a neural controller, implemented using FiLM layers to achieve this conditioning; trains a quadrotor control policy via model-free reinforcement learning to navigate through a series of waypoints in the shortest time possible, with adjustable aggressiveness during deployment.
Using a Collaborative Robotic Arm as Human-Machine Interface: System Setup and Application to Pose Control Tasks
C. Braun, S. Hohmann
OptimizationRobotic Intelligence
🎯 What it does: This paper describes a system setup using the KUKA LBR iiwa 14 R820 collaborative robot as a human-machine interface (HMI), and determines the optimal initial posture through Yoshikawa's manipulability measure analysis. Subsequently, comparative experiments validate its performance against existing high-end tactile HMIs, and its advantages are demonstrated in two planetary exploration applications (remote-controlled planetary rover planar turntable and shared control with a robot).
Using Learning Curve Predictions to Learn from Incorrect Feedback
Taylor A. Kessler Faulkner, A. Thomaz
Reinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: Developed the CLEAR algorithm, which uses reinforcement learning learning curves to filter out incorrect human feedback, enabling learning agents to learn even when teacher information is imperfect.
Using Memory-Based Learning to Solve Tasks with State-Action Constraints
Mrinal Verghese, C. Atkeson
Reinforcement Learning
🎯 What it does: Propose a memory-based learning method to address state-action constraint tasks, evaluated in both real and simulated environments.
Using Registration with Fourier-SOFT in 2D (FS2D) for Robust Scan Matching of Sonar Range Data
Tim Hansen, Andreas Birk
Robotic IntelligenceSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed and implemented Fourier-SOFT 2D (FS2D), a new robust scan matching method for underwater sonar range data.