ICRA 2025 Papers — Page 15
IEEE International Conference on Robotics and Automation · 1604 papers
SuPLE: Robot Learning with Lyapunov Rewards
P. Nguyễn, Stas Tiomkin
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a reward function SuPLE based on the system's Lyapunov exponents, construct a computational framework, and verify its effectiveness in stabilization tasks for classical dynamical systems.
Surface Roughness Estimation for Terrain Perception
Minxiang Ye, Anhuan Xie
Convolutional Neural NetworkImageMultimodality
🎯 What it does: Propose a new task of estimating pixel-level terrain roughness using RGB images, and provide semantic-aware and edge-aware roughness descriptors.
Surfaceaug: Toward Versatile, Multimodally Consistent Ground Truth Sampling
Ryan Rubel, Andrew Dudash
Object DetectionData SynthesisAutonomous DrivingImagePoint Cloud
🎯 What it does: Developed a ground truth sampling algorithm called SurfaceAug, which achieves target-level transformations in both modalities by resampling and pasting targets in images and point clouds.
SurgPLAN++: Universal Surgical Phase Localization Network for Online and Offline Inference
Zhen Chen, Hongbin Liu
ClassificationSegmentationVideoBiomedical Data
🎯 What it does: A universal network called SurgPLAN++ was designed for identifying phases during online and offline surgical stages, achieving whole-video phase segmentation prediction through phase proposals based on the time detection principle.
SurgPose: a Dataset for Articulated Robotic Surgical Tool Pose Estimation and Tracking
Zijian Wu, S. Salcudean
Object TrackingPose EstimationImageBiomedical DataBenchmark
🎯 What it does: This paper annotates instantiated semantic keypoints by using UV-reactive paint on surgical robot instruments, constructing a dataset named SurgPose. It collects approximately 120,000 instances of surgical instruments and uses stereo cameras to lift 2D keypoints into 3D.
SurgPose: Generalisable Surgical Instrument Pose Estimation Using Zero-Shot Learning and Stereo Vision
Utsav Rai, S. Giannarou
Pose EstimationDepth EstimationConvolutional Neural NetworkSupervised Fine-TuningOptical FlowImageBenchmark
🎯 What it does: Proposed a 6 degrees of freedom (6DoF) surgical tool pose estimation pipeline using zero-shot learning and stereo vision, performing depth estimation with RAFT-Stereo and replacing SAM's instance segmentation module with a fine-tuned Mask R-CNN;
Suture Thread Modeling Using Control Barrier Functions for Autonomous Surgery
Kimia Forghani, Y. Diaz-Mercado
Computational EfficiencyRobotic Intelligence
🎯 What it does: Proposed and implemented a scheme that models the dynamics of a suture needle thread using Control Barrier Functions (CBFs) and Control Lyapunov Functions (CLFs), achieving a thread model that is both realistic and computationally efficient.
Swept Volume-Aware Trajectory Planning and MPC Tracking for Multi-Axle Swerve-Drive AMRs
Tianxin Hu, Lihua Xie
Autonomous DrivingOptimization
🎯 What it does: Designed a sweep volume-aware trajectory planning and MPC tracking framework for multi-axis arm-driven AMRs, capable of real-time minimization of sweep volume and maintaining precise control during turns.
SYNERGAI: Perception Alignment for Human-Robot Collaboration
Yixin Chen, Siyuan Huang
Robotic IntelligenceGraph Neural NetworkTransformerLarge Language ModelMultimodalityGraph
🎯 What it does: Developed the SYNERGAI system to achieve perceptual alignment and human-robot collaboration.
SynerGuard: A Robust Framework for Point Cloud Classification via Local Geometry and Spatial Topology
Haonan Zhong, Yang Song
ClassificationAdversarial AttackPoint Cloud
🎯 What it does: Proposes the SynerGuard framework, leveraging local geometric and spatial topological features to enhance the robustness of point cloud classification models against adversarial attacks;
Synthesizing Depowdering Trajectories for Robot Arms using Deep Reinforcement Learning
Maximilian Maurer, Shahram Eivazi
Robotic IntelligenceReinforcement Learning
🎯 What it does: Researched and implemented trajectory synthesis for a robotic arm based on deep reinforcement learning, utilizing compressed air spray to remove powder from different 3D object surfaces, and developed GPU-accelerated vectorized cleaning effect simulation as well as UV mapping-based visual-free trajectory synthesis.
Synthesizing Grasps and Regrasps for Complex Manipulation Tasks
Aditya Patankar, Nilanjan Chakraborty
Robotic IntelligencePoint Cloud
🎯 What it does: Proposed a grasping and re-grasping algorithm for complex manipulation tasks based on point cloud data, achieving the target through feasible regions and continuous constant helical motion sequences.
System-Level Safety Monitoring and Recovery for Perception Failures in Autonomous Vehicles
Kaustav Chakraborty, Somil Bansal
Autonomous DrivingReinforcement LearningPoint Cloud
🎯 What it does: Developed a Q-network called SPARQ to assess the safety of planning strategies in real-time during autonomous driving system operations, and provide corrective measures when perception failures lead to potential safety risks.
Systematic Comparison of Projection Methods for Monocular 3D Human Pose Estimation on Fisheye Images
Stephanie Käs, Bastian Leibe
Pose EstimationImageBenchmark
🎯 What it does: Evaluated the effectiveness of pinhole, equidistant, dual-sphere, and cylindrical projection models for 3D human pose estimation in monocular fisheye images
TacDiffusion: Force-Domain Diffusion Policy for Precise Tactile Manipulation
Yansong Wu, Sami Haddadin
Robotic IntelligenceDiffusion model
🎯 What it does: Propose a framework based on diffusion models to generate 6D torque for accomplishing high-precision tactile insertion tasks.
Tactile Functasets: Neural Implicit Representations of Tactile Datasets
Sikai Li, Nima Fazeli
Pose EstimationRepresentation Learning
🎯 What it does: Proposes a neural implicit function representation for compressing and reconstructing high-dimensional raw images from tactile sensors, validating its effectiveness in the object pose estimation task when grasping an object.
Take Your Best Shot: Sampling-Based Planning for Autonomous Photography
Shijie Gao, N. Bezzo
OptimizationRobotic IntelligenceImage
🎯 What it does: Proposed a sampling-based planning framework for autonomous photography, selecting the optimal shooting viewpoint through evaluation metrics and derivative-free optimization methods.
Talk2Radar: Bridging Natural Language with 4D mmWave Radar for 3D Referring Expression Comprehension
Runwei Guan, Hui Xiong
RecognitionConvolutional Neural NetworkGraph Neural NetworkTextMultimodalityPoint Cloud
🎯 What it does: Constructed the first Talk2Radar dataset, integrating 4D mmWave radar point clouds with natural language prompts for 3D gesture expression understanding, and proposed the T-RadarNet model.
TANGO: Traversability-Aware Navigation with Local Metric Control for Topological Goals
Stefan Podgorski, Ian Reid
Depth EstimationRobotic IntelligenceImage
🎯 What it does: Propose a topology-metric navigation pipeline based on RGB and object-level targets, achieving zero-shot long-term robot navigation without requiring 3D maps or pre-trained controllers.
Target-Aware Viewpoint Generation for Active Robotic Exploration in Unknown Environments
Pu Xu, Zheng Fang
OptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Proposed a target-aware robot exploration framework, including lightweight 3D object detection, perspective generation based on information gain and inspection gain, and hierarchical active exploration based on heuristic methods, aiming to enhance target search efficiency and completeness.
Targeted Parallelization of Conflict-Based Search for Multi-Robot Path Planning
Teng Guo, Jingjin Yu
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Researched and implemented parallelization strategies for conflict search, including distributed parallel branch exploration for small dense instances, and node expansion priority and conflict prioritization resolution for large sparse instances.
Task-Aware Semantic Map: Autonomous Robot Task Assignment Beyond Commands
Daewon Choi, Yoonseon Oh
Robotic IntelligenceTransformerLarge Language ModelImage
🎯 What it does: Proposed a new semantic map called Task-Aware Semantic Map (TASMap), enabling robots to autonomously allocate and suggest necessary tasks in a scene without explicit human instructions.
Task-Oriented 6-DoF Grasp Pose Detection in Clutters
An-Lan Wang, Wei-Shi Zheng
Pose EstimationRobotic IntelligenceImagePoint Cloud
🎯 What it does: Studied task-oriented 6-DoF grasp pose detection in cluttered scenes and constructed a corresponding dataset.
Task-Oriented Pre-Training for Drivable Area Detection
Fulong Ma, Jun Ma
Autonomous DrivingSupervised Fine-TuningVision Language ModelContrastive LearningImage
🎯 What it does: Propose a task-oriented pre-training method that first uses Segment Anything (SAM) to generate redundant segmentation candidates, then fine-tunes Contrastive Language-Image Pre-training (CLIP) with Specific Category Enhancement Fine-tuning (SCEF) to select candidates most relevant to drivable areas, generating coarse-grained training data and further refining the model on manually annotated data.
Task-Specific Embodied Tactile Sensing for Dexterous Hand
Qi Wei, Qiang Li
Computational EfficiencyRobotic Intelligence
🎯 What it does: Propose an ET-Hand hand that realizes a task-specific multimodal sensor placement framework and dynamically adjusts the position, type, and quantity of tactile sensors to reduce sensor resource waste.
TaskExp: Enhancing Generalization of Multi-Robot Exploration with Multi-Task Pre-Training
Shaohao Zhu, Jinming Xu
Robotic IntelligenceReinforcement Learning
🎯 What it does: Developed a multi-task pre-training algorithm called TaskExp to enhance the generalization performance of multi-robot exploration;
TCAFF: Temporal Consistency for Robot Frame Alignment
Mason B. Peterson, Jonathan P. How
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: A multi-hypothesis algorithm named TCAFF was developed for aligning coordinate frames of neighboring robots, and its effectiveness was verified in an experiment involving four robots collaboratively tracking six pedestrians.
TDFANet: Encoding Sequential 4D Radar Point Clouds Using Trajectory-Guided Deformable Feature Aggregation for Place Recognition
Shouyi Lu, Qiang Shu
RecognitionConvolutional Neural NetworkPoint CloudSequential
🎯 What it does: Proposed a trajectory-guided deformable feature aggregation network (TDFANet) based on 4D radar sequence point clouds for place recognition.
Teleoperating a 6 DoF Robotic Manipulator from Head Movements
Alexis Poignant, Guillaume Morel
Robotic Intelligence
🎯 What it does: Proposed an interactive control method for remotely operating a nearby 6-DOF robotic manipulator using only head movements
Tendon Locking for Antagonistic Configuration- and Stiffness-Control in Soft Robots
J. Licher, Helge A. Wurdemann
Robotic Intelligence
🎯 What it does: Propose an antagonistic stiffening mechanism combining pneumatic actuation and tendon locking to achieve controllability in configuration and stiffness for soft robots.
Tension Dependent Twisted String Actuator Modelling and Efficacy Benchmarking in Force and Impedance Control
Christopher Herneth, Sami Haddadin
BenchmarkPhysics Related
🎯 What it does: Conduct comprehensive experimental analysis of torsional spring actuators (TSA), improve the accuracy of contraction modeling, and establish benchmarks for tension and impedance control; propose a torsion radius function based on axial tension and verify it experimentally; evaluate the accuracy, precision, impact stability, and bandwidth of tension control; investigate the performance of joint impedance control in disturbance stability and position control bandwidth.
Tensiworm: A Novel Tensegrity Robot with Enhanced Peristaltic Locomotion Efficiency
Christian Kazoleas, Sichen Yuan
Robotic Intelligence
🎯 What it does: Proposed and implemented a three-unit tension-balanced robot named Tensiworm, which achieves earthquake-worm-like forward locomotion using active cables driven by shape memory materials.
Terrain-Aware Model Predictive Control of Heterogeneous Bipedal and Aerial Robot Coordination for Search and Rescue Tasks
Abdulaziz Shamsah, Ye Zhao
OptimizationRobotic Intelligence
🎯 What it does: Designed a task and motion planning framework for a heterogeneous team of bipedal robots and drones in search and rescue missions.
Text2Robot: Evolutionary Robot Design from Text Descriptions
Ryan P. Ringel, Boyuan Chen
GenerationOptimizationRobotic IntelligenceTextMesh
🎯 What it does: This paper proposes the Text2Robot framework, which can convert users' text specifications and performance preferences into manufacturable quadruped robots. It rapidly generates diverse morphologies through text-to-3D model technology within minutes, followed by generating walkable robots through geometric processing algorithms and body control co-optimization within a day.
THAMP-3D: Tangent-Based Hybrid A* Motion Planning for Tethered Robots in Sloped 3D Terrains
Rahul Kumar, Sze Zheng Yong
Robotic Intelligence
🎯 What it does: A novel motion planning algorithm is developed for path planning of curvature-constrained cable robot teams operating on tilted three-dimensional terrains.
The Automation of Uncrewed Aircraft Systems Traffic Management Calibration Based on Experimental Platform Data
T. Henderson, William Raley
Autonomous Driving
🎯 What it does: Experiments were conducted on a test platform, proposing and demonstrating a method to calibrate the safety parameters of the Unmanned Traffic Management (UTM) system based on the operational characteristics of the platform.
The Devil is in the Quality: Exploring Informative Samples for Semi-Supervised Monocular 3D Object Detection
Zhipeng Zhang, Heng Fan
Object DetectionImage
🎯 What it does: Proposes a general framework called 'Augment and Criticize' to address the problem of insufficient information samples in semi-supervised monocular 3D object detection, utilizing APG to generate pseudo labels and CRS to evaluate their contribution.
The Experiment Orchestration System (EOS): Comprehensive Foundation for Laboratory Automation
Angelos Angelopoulos, Ron Alterovitz
🎯 What it does: Proposed and implemented an open-source experimental scheduling system called EOS, providing an extensible framework that supports defining laboratories, devices, tasks, experiments, and optimization criteria using YAML and Python plugins, and controlling experimental devices through distributed runtime and a central coordinator.
The Impact of Sensor Faults on Connected Autonomous Vehicle Localization
Shinsaku Kuwada, Matthew Spenko
Autonomous DrivingSimultaneous Localization and Mapping
🎯 What it does: Developed and evaluated an integrity monitoring method for cooperative localization in connected autonomous vehicles to analyze the impact of sensor failures on localization safety.
The Influence of Counterbalance System on the Dynamic Characterization of Heavy Industrial Robots
Julen Urrutia, Jon Larrañaga
Robotic IntelligencePhysics Related
🎯 What it does: Developed an inertial parameter estimation method for heavy industrial robots, taking into account the influence of the balance system.
The Ingredients for Robotic Diffusion Transformers
S. Dasari, Sergey Levine
Robotic IntelligenceTransformerDiffusion modelMultimodality
🎯 What it does: Identify, study, and improve key architectural design decisions for high-capacity diffusion Transformers, proposing the efficient DiT-Block Policy that solves various robotic tasks without tedious hyperparameter tuning.
The Mini Wheelbot: A Testbed for Learning-based Balancing, Flips, and Articulated Driving
Henrik Hose, Sebastian Trimpe
OptimizationHyperparameter SearchRobotic Intelligence
🎯 What it does: Designed and implemented Mini Wheelbot—a balancing inverted sub-wheel robot capable of self-standing from any initial posture, and tested two learning-based control algorithms on it.
The qPCRBot: Combining Automated Data Handling, Standardization, and Robotic Labware Transport for Better qPCR Measurements
Henning Zwirnmann, Sami Haddadin
Robotic Intelligence
🎯 What it does: Developed the qPCRBot system to enhance the efficiency and reproducibility of qPCR experiments through automated data processing, standardized management, and robotic transportation of experimental labware.
The Radiance of Neural Fields: Democratizing Photorealistic and Dynamic Robotic Simulation
Georgina Nuthall (University of Surrey), Oscar Alejandro Mendez Maldonado (University of Surrey)
Robotic IntelligenceNeural Radiance FieldMultimodality
🎯 What it does: Developed an integrated robot simulator combining dual NeRF neural rendering, neural animated human entities, and multi-sensor outputs, enabling realistic visual rendering and human-robot interaction in dynamic environments;
The Role of Tactile Sensing for Learning Reach and Grasp
Boya Zhang, G. Martius
Robotic IntelligenceReinforcement Learning
🎯 What it does: Compared various tactile and environmental setups using two model-free reinforcement learning methods for adversarial grasping
The Spinning Blimp: Design and Control of a Novel Minimalist Aerial Vehicle Leveraging Rotational Dynamics and Locomotion
Leonardo Santens, David Saldaña
Robotic IntelligencePhysics Related
🎯 What it does: Designed and controlled a minimalist lighter-than-air vehicle called Spinning Blimp, utilizing rotational dynamics and maneuverability to achieve low-energy stable flight.
ThermoStereoRT: Thermal Stereo Matching in Real Time via Knowledge Distillation and Attention-Based Refinement
Anning Hu, Danping Zou
Depth EstimationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Designed and implemented ThermoStereoRT, a real-time thermal imaging stereo matching method that constructs a 3D cost volume using a lightweight backbone network, generates initial disparity maps through a multi-scale attention mechanism, and refines them using channel and spatial attention modules; improves performance on sparsely annotated thermal imaging data via knowledge distillation while maintaining real-time speed; applicable to all-weather scenarios such as nighttime drone surveillance or bed-under cleaning robots.
Think Deep and Fast: Learning Neural Nonlinear Opinion Dynamics from Inverse Dynamic Games for Split-Second Interactions
Haimin Hu, J. F. Fisac
Autonomous DrivingReinforcement Learning
🎯 What it does: Propose a Neural Nonlinear Opinion Dynamics (Neural NOD) model based on learning and game theory, which automatically learns parameters from expert demonstrations to achieve fast deadlock-free decision-making.
Think on Your Feet: Seamless Transition Between Human-Like Locomotion in Response to Changing Commands
Huaxing Huang, Zheyuan Jiang
Robotic IntelligenceReinforcement LearningGenerative Adversarial Network
🎯 What it does: This paper proposes a method that combines human-like motion transfer with precise velocity tracking, enabling robots to achieve seamless multi-posture gait transitions under continuously changing commands, and supporting zero-shot transfer in both simulation and real-world environments;
This&That: Language-Gesture Controlled Video Generation for Robot Planning
Boyang Wang, J. Park
GenerationRobotic IntelligenceVision-Language-Action ModelDiffusion modelVideoMultimodality
🎯 What it does: Proposed the This&That framework, combining language and gesture-controlled video generation for communication, planning, and execution in robotic tasks.
Three-Dimension Tip Force Perception and Axial Contact Location Identification for Flexible Endoscopy Using Tissue-Compliant Soft Distal Attachment Cap Sensors
Tao Zhang, Hongliang Ren
MultimodalityBiomedical Data
🎯 What it does: Designed a flexible, tissue-compatible soft tip attachment sensor cap based on Fiber Bragg Grating (FBG) to achieve terminal surface contact position identification and three-dimensional contact force perception; the sensor cap can be directly installed on the tip of standard endoscopes, helping operators adjust the bending section during passage through natural cavities based on contact information; experimental validation through finite element analysis simulation and learning-based calibration process demonstrates contact force error below 3% and contact position identification accuracy reaching 98.8%.
Through the Clutter: Exploring the Impact of Complex Environments on the Legibility of Robot Motion
Melanie Schmidt-Wolf, David Feil-Seifer
Robotic Intelligence
🎯 What it does: Proposed an entropy-based environmental chaos metric and a potential field-based robot motion planner, verifying its readability in a cluttered environment for sorting tasks in a simulation tool
Tightly Coupled Range Inertial Odometry and Mapping with Exact Point Cloud Downsampling
Kenji Koide, Masashi Yokozuka
Autonomous DrivingOptimizationComputational EfficiencySimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed a point cloud downsampling algorithm based on coreset extraction, and constructed a complete SLAM framework, including sliding-window-based odometry estimation and global trajectory optimization that minimizes full-graph registration error, all of which can be implemented in real-time on standard CPUs.
Time-Correlated Model Predictive Path Integral: Smooth Action Generation for Sampling-Based Control
Minhyeong Lee, Dongjun Lee
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Propose the Time-Correlated Model Predictive Path Integral (TC-MPPI) method, which directly introduces the temporal correlation of actions into the sampled control to reduce action noise.
TinySense: A Lighter Weight and More Power-Efficient Avionics System for Flying Insect-Scale Robots
Zhitao Yu, Sawyer B. Fuller
OptimizationComputational EfficiencyRobotic IntelligenceOptical Flow
🎯 What it does: This study improved the sensor suite of a drone weighing less than one gram (FIR), further reducing the aircraft's weight and power consumption, and achieved hover control through sensor fusion.
To Ask or not to Ask: Human-in-the-loop Contextual Bandits with Applications in Robot-Assisted Feeding
Rohan Banerjee, T. Bhattacharjee
Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningBiomedical Data
🎯 What it does: Proposes a context-aware multi-armed bandit framework called LINUCB-QG, which predicts query workload to decide whether to request user feedback, balancing task performance and user workload.
Tool-Mediated Robot Perception of Granular Substances Using Multiple Sensory Modalities
Sishi Liu, Jivko Sinapov
Robotic IntelligenceMultimodalityAudio
🎯 What it does: Proposes a multi-perception robot framework that utilizes various tools (e.g., spoons, forks) and exploration behaviors (e.g., stirring, probing) to collect non-visual sensing data such as audio, tactile, and haptic information, and perceives the properties of granular materials through time window segmentation and modal alignment.
Topological Mapping for Traversability-Aware Long-Range Navigation in Off-Road Terrain
J. Tremblay, D. Meger
Autonomous DrivingTransformerSimultaneous Localization and MappingImage
🎯 What it does: Propose a method based on topological graphs for end-to-end planning, exploration, and low-level control, utilizing vision and GPS to achieve long-distance navigation in unknown off-road forest terrains.
Topology-Based Visual Active Room Segmentation
Chenyu Bao, T. Lam
SegmentationSimultaneous Localization and MappingImage
🎯 What it does: This paper proposes a proactive room segmentation framework that can incrementally and autonomously complete room segmentation in cluttered indoor environments.
Toward Zero-Shot Learning for Visual Dehazing of Urological Surgical Robots
Renkai Wu, Hao Tang
RestorationRobotic IntelligenceImageBiomedical DataBenchmark
🎯 What it does: Proposed an unsupervised zero-shot dehazing method named RSF-Dehaze, and created the USRobot-Dehaze dataset for urological robotic vision.
Towards Accurate Semi-Supervised BEV 3D Object Detection with Depth-Aware Refinement and Denoising-Aided Alignment
Zhao Yang, Longjun Liu
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: Proposes a semi-supervised BEV 3D object detection framework based on a small amount of labeled data and a large amount of unlabeled data, and designs depth self-refinement, denoising label regression, and consistency target guided alignment modules.
Towards Autonomous Data Annotation and System-Agnostic Robotic Grasping Benchmarking with 3D-Printed Fixtures
W. Boerdijk, Rudolph Triebel
Pose EstimationRobotic IntelligenceSupervised Fine-TuningImageBenchmark
🎯 What it does: This paper proposes the use of 3D-printed fixtures designed for any rigid object to temporarily fix objects in the scene, extract their poses, and then remove the fixtures to maintain a natural scene and achieve automated data annotation.
Towards Autonomous Verification: Integrating Cognitive AI and Semantic Digital Twins in Medical Robotics
Patrick Mania, Michael Beetz
Robotic IntelligenceWorld ModelBiomedical Data
🎯 What it does: Proposes an imagination-driven perception framework that integrates cognitive AI with semantic digital twins, enabling medical robots to simulate task outcomes, compare them with reality, and automatically verify the success of actions.
Towards Autonomous Wood-Log Grasping with a Forestry Crane: Simulator and Benchmarking
M. Vu (Vienna University of Technology), Andreas Kugi (Vienna University of Technology)
Robotic IntelligenceReinforcement LearningBenchmarkAgriculture Related
🎯 What it does: Explored the feasibility of using reinforcement learning to achieve autonomous grasping and lifting of wooden logs by forestry cranes, constructed a simulator based on Mujoco, and implemented a velocity controller.
Towards Closing the Loop in Robotic Pollination for Indoor Farming Via Autonomous Microscopic Inspection
Chuizheng Kong, Shreyas Kousik
Object DetectionPose EstimationRobotic IntelligenceImageAgriculture Related
🎯 What it does: Designed and implemented a robot system for indoor farms, including a 7-degree-of-freedom robotic arm and a custom end-effector equipped with an endoscope camera, a 2-degree-of-freedom microscope system, and a vibration pollination tool, used for detecting, locating strawberry flowers, navigating, pollinating, and inspecting through the microscope.
Towards Effective Utilization of Mixed-Quality Demonstrations in Robotic Manipulation via Segment-Level Selection and Optimization
Jingjing Chen, Cewu Lu
OptimizationRobotic IntelligenceContrastive Learning
🎯 What it does: Proposes the S2I framework, which improves the utilization efficiency of mixed-quality demonstration data through segment-level segmentation, contrastive learning selection, and trajectory optimization
Towards Evaluating the User Comfort and Experience of a Novel Steerable Drilling Robotic System in Pedicle Screw Fixation Procedures: A User Study
Susheela Sharma, F. Alambeigi
Robotic Intelligence
🎯 What it does: Studied and evaluated the user comfort and experience of a collaborative adjustable drilling robot system in pedicle screw insertion surgery.
Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-Guided 3D Policy
Ricardo Garcia, Cordelia Schmid
Robotic IntelligenceLarge Language ModelVision Language ModelVision-Language-Action ModelMultimodalityBenchmark
🎯 What it does: Propose the GemBench benchmark to evaluate the generalization ability of robot manipulation strategies under visual-language conditions, and assess existing state-of-the-art methods on this benchmark; propose the 3D-LOTUS method that leverages rich 3D information for action prediction, and further introduce the 3D-LOTUS++ framework that integrates the motion planning of 3D-LOTUS with the task planning capability of LLMs and the object localization ability of VLMs to enhance generalization performance on new tasks.
Towards Latency-Aware 3D Streaming Perception for Autonomous Driving
Jiaqi Peng, Yuan Shen
Autonomous DrivingComputational EfficiencyPoint CloudBenchmark
🎯 What it does: Proposes a latency-aware framework and evaluation benchmark for real-time 3D streaming perception
Towards Neurorobotic Interface for Finger Joint Angle Estimation: A Multi-Stage CNN-LSTM Network with Transfer Learning
Yun Chen, Qiang Zhang
Pose EstimationRobotic IntelligenceConvolutional Neural NetworkRecurrent Neural NetworkMultimodalityBiomedical DataUltrasound
🎯 What it does: Propose a multi-stage cascaded CNN-LSTM network combined with an upsampling module to enhance multimodal finger joint angle estimation using forearm sEMG and ultrasound images;
Towards Open-Ended Robotic Exploration Using Vision-Inspired Similarity and Foundation Models
P. Filntisis, Petros Maragos
Robotic IntelligenceTransformerVision-Language-Action ModelContrastive LearningImage
🎯 What it does: Propose the VISOR framework, which leverages visual similarity and foundational models to enable autonomous exploration and learning for robots, capable of classifying encountered objects into known, unknown, distractors, or text-specified objects, and performing scene segmentation and feature extraction without training;
Towards Over-Canopy Autonomous Navigation: Crop-Agnostic LiDAR-Based Crop-Row Detection in Arable Fields
Ruiji Liu, George Kantor
Autonomous DrivingPoint CloudAgriculture Related
🎯 What it does: Proposes a LiDAR-based crop row detection and navigation system that enables crop-agnostic autonomous navigation even when the vegetation canopy completely obscures the row spacing, capable of detecting row ends and automatically switching to the next row;
Towards Perpetually-Deployable Ubiquitous Aerial Robotics: An Amphibious Self-Sustainable Solar Small-UAS
S. Carlson, Christos Papachristos
Robotic IntelligenceImage
🎯 What it does: Developed a waterproof small drone named Gannet Solar-VTOL, capable of prolonged stay on water surfaces, self-sustaining through solar charging, and entering low-power sleep mode during nighttime or insufficient lighting conditions, while integrating a camera and neural processing unit for on-site environmental monitoring.
Towards Real-Time Generation of Delay-Compensated Video Feeds for Outdoor Mobile Robot Teleoperation
Neeloy Chakraborty, K. Driggs-Campbell
Robotic IntelligenceVideoAgriculture Related
🎯 What it does: Proposed a modular learning-based visual pipeline that real-time generates delay-compensated images to enhance the quality of monitoring videos during remote control of agricultural robots.
Towards Robust Autonomous Driving: Conditional Multimodal Large Language Models for Fine-Grained Perception
Fengzhao Sun, Fang Gao
Autonomous DrivingLarge Language ModelVision Language ModelMultimodality
🎯 What it does: Proposes Percept-DriveLM, a multimodal large language model designed for fine-grained perception tasks in autonomous driving, and develops a core visual fusion module;
Towards Safe and Efficient Through-the-Canopy Autonomous Fruit Counting with UAVs
Teaya Yang, Mark W. Mueller
Object DetectionAutonomous DrivingSimultaneous Localization and MappingImageAgriculture Related
🎯 What it does: Proposed a safe and efficient autonomous drone system for fruit counting by flying through tree canopies
Towards Safe and Energy-Efficient Real-Time Motion Planning in Windy Urban Environments
Spencer Folk, Vijay Kumar
Autonomous DrivingOptimization
🎯 What it does: Propose a framework based on onboard sensing for local wind field prediction and receding horizon optimal control, achieving safe and energy-efficient navigation for low-altitude drones in urban wind fields.
Towards Survivability in Complex Motion Scenarios: RGB-Event Object Tracking via Historical Trajectory Prompting
Wenhao Xia, Xu Jia
Object TrackingPrompt EngineeringMultimodalityBenchmark
🎯 What it does: Propose the EventTPT framework, leveraging key hints from historical trajectories to enhance RGB-Event object tracking performance.
Towards Transparent Multi-Agent Autonomous Systems Through Principled Multi-Source Knowledge Distillation
Zhong Guo, Wen Yao
Explainability and InterpretabilityKnowledge DistillationReinforcement LearningGraph
🎯 What it does: Proposed an interpretable multi-agent path finding framework that uses behavior trees trained with reinforcement learning to simulate multi-agent path finding tasks
Tracking Everything in Robotic-Assisted Surgery
Bohan Zhan, Baoru Huang
Object TrackingRobotic IntelligenceVideoBiomedical DataBenchmark
🎯 What it does: A specialized video tracking dataset for robotic-assisted surgery was constructed, existing Tracking Any Point (TAP) methods were evaluated on this dataset, and a new tracking algorithm named SurgMotion was proposed to enhance tracking performance.
TrackOcc: Camera-Based 4D Panoptic Occupancy Tracking
Zhuoguang Chen, Hang Zhao
Object TrackingSegmentationAutonomous DrivingVideo
🎯 What it does: Proposed a 4D panoramic occupancy tracking task based on cameras, and introduced the end-to-end TrackOcc method;
Traffic Regulation-aware Path Planning with Regulation Databases and Vision-Language Models
Xu Han, Jiaqi Ma
Autonomous DrivingVision Language ModelImageTextMultimodality
🎯 What it does: Integrating traffic regulation compliance into autonomous driving systems by utilizing RGB cameras and visual-language models to generate descriptive text, combined with a readable regulation database to support rule-aware path planning.
Training Human-Robot Teams by Improving Transparency Through a Virtual Spectator Interface
Sean Dallas, D. Tilbury
Explainability and InterpretabilityRobotic Intelligence
🎯 What it does: Proposed and evaluated a training review tool called Virtual Spectator Interface (VSI) to enhance the performance and situational awareness of human-robot teams in simulated search tasks.
Trajectory Planning and Control for Differentially Flat Fixed-Wing Aerial Systems
Luca Morando, Giuseppe Loianno
OptimizationRobotic Intelligence
🎯 What it does: Proposed a fast and efficient real-time trajectory planning and control method for fixed-wing UAVs.
Trajectory Planning with Signal Temporal Logic Costs Using Deterministic Path Integral Optimization
Patrick Halder, Matthias Althoff
Autonomous DrivingOptimizationBenchmark
🎯 What it does: Propose a sampling method based on model predictive path integral control to solve optimal control problems with signal temporal logic cost functions.
TrajSSL: Trajectory-Enhanced Semi-Supervised 3D Object Detection
P. Jacobson, Ming C. Wu
Object DetectionAutonomous DrivingVideoPoint Cloud
🎯 What it does: Generate trajectories using a pre-trained motion prediction model, combine multi-frame consistency to enhance pseudo label quality, and directly insert predicted trajectories into the scene to compensate for missed detections.
TransDiff: Diffusion-Based Method for Manipulating Transparent Objects Using a Single RGB-D Image
Haoxiao Wang, Hao Dong
Depth EstimationDiffusion modelImage
🎯 What it does: Proposes TransDiff, a depth completion framework for transparent objects based on single-view RGB-D, utilizing DDPM to achieve object grasping.
Transferring Visual Knowledge: Semi-Supervised Instance Segmentation for Object Navigation Across Varying Height Viewpoints
Qiu Zheng, T. Lam
SegmentationDomain AdaptationContrastive LearningImage
🎯 What it does: Propose a semi-supervised instance segmentation method for target navigation under different camera height perspectives, reducing the need for additional annotations by transferring knowledge from the source height to the target height.
TransForce: Transferable Force Prediction for Vision-Based Tactile Sensors with Sequential Image Translation
Zhuo Chen, Shan Luo
Image TranslationDomain AdaptationRecurrent Neural NetworkImageSequential
🎯 What it does: Proposed TransForce, a transferable force prediction model that utilizes image-force paired data to achieve sequence image translation for visual tactile sensors and recursive force estimation, enhancing force prediction accuracy of new sensors under different lighting colors and marking patterns.
TransformerMPC: Accelerating Model Predictive Control via Transformers
Vrushabh Zinage, E. Bakolas
OptimizationComputational EfficiencyTransformer
🎯 What it does: Propose the TransformerMPC method, leveraging the Transformer attention mechanism to achieve online constraint removal and better warm start initialization, thereby accelerating MPC computation.
TranSplat: Surface Embedding-Guided 3D Gaussian Splatting for Transparent Object Manipulation
Jeong-Man Kim, Ayoung Kim
Depth EstimationDiffusion modelGaussian SplattingImageBenchmark
🎯 What it does: Propose a surface embedding guided 3D Gaussian splat (TranSplat) method for transparent objects to achieve accurate and dense depth completion.
TREND: Tri-Teaching for Robust Preference-based Reinforcement Learning with Demonstrations
Shuaiyi Huang, Abhinav Shrivastava
Reinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: By integrating a small number of expert demonstrations with a tri-teaching strategy, three reward models are trained to counteract noisy preference feedback, enhancing the robustness of preference-based reinforcement learning.
Tri-AutoAug: Single Domain Generalization for Bird's-Eye-View 3D Object Detection Through Pixel-2D-3D Features
Xue Zhao, Nanyang Ye
Domain AdaptationAutonomous DrivingImagePoint Cloud
🎯 What it does: Propose the three-layer automatic augmentation (Tri-AutoAug) method for single-domain generalization (SDG) bird's-eye view (BEV) 3D object detection, aiming to learn more domain-invariant features through diversified enhancement of images and 2D features;
TriHRCBot: A Robotic Architecture for Triadic Human-Robot Collaboration Through Mediated Object Alignment
Francesco Semeraro, Angelo Cangelosi
Robotic Intelligence
🎯 What it does: This paper proposes the TriHRCBot architecture, enabling two parallel users to simultaneously manipulate a target object through robot perception and pose adjustment.
TrofyBot: A Transformable Rolling and Flying Robot with High Energy Efficiency
Mingwei Lai, Yanjun Cao
Robotic Intelligence
🎯 What it does: Proposed and implemented a deformable rolling and flying robot named TrofyBot, focusing on enhancing energy efficiency and mode switching.
Trustworthy Robot Behavior Tree Generation Based on Multi-Source Heterogeneous Knowledge Graph
Jianchao Yuan, Jianping Tang
GenerationRobotic IntelligenceGraph
🎯 What it does: Propose a method for automatically generating trusted behavior trees based on multi-source, heterogeneous, high-quality robot knowledge graphs, automating the generation of plan-level behavior tree structures.
TS-DETR: Traffic Sign Detection Based on Positive and Negative Sample Augmentation
Ching-Lung Lin, Chieh-Chih Wang
Object DetectionTransformerImage
🎯 What it does: Proposes an end-to-end traffic sign detection framework based on DETR, and introduces positive and negative sample enhancement along with the UASPP module to improve recognition performance.
TSCLIP: Robust CLIP Fine-Tuning for Worldwide Cross-Regional Traffic Sign Recognition
Guoyang Zhao, Jun Ma
RecognitionSupervised Fine-TuningPrompt EngineeringVision Language ModelImageBenchmark
🎯 What it does: Propose TSCLIP, a robust fine-tuning method based on CLIP for global cross-regional traffic sign recognition, and construct a cross-regional traffic sign benchmark dataset, adopting prompt engineering schemes tailored to traffic sign characteristics and Adaptive Dynamic Weight Ensembling (ADWE) technology.
TSPDiffuser: Diffusion Models as Learned Samplers for Traveling Salesperson Path Planning Problems
Ryo Yonetani
OptimizationDiffusion model
🎯 What it does: Proposes TSPDiffuser, which employs a learning sampler trained with diffusion models to generate the shortest collision-free paths for the Traveling Salesman Path Planning problem in obstacle-filled environments.
Tunable Leg Stiffness in a Monopedal Hopper for Energy-Efficient Vertical Hopping Across Varying Ground Profiles
Rongqian Chen, Wei-Hsi Chen
OptimizationRobotic Intelligence
🎯 What it does: Designed and implemented HASTA—a vertical jumping robot with real-time adjustable foot stiffness—and determined through experiments and simulations the optimal foot stiffness under different ground stiffness and damping conditions to maximize steady-state jump height.