ICRA 2025 Papers — Page 14
IEEE International Conference on Robotics and Automation · 1604 papers
SC-Former: A Segmentation Convolution Transformer for Lung Surgery Robots
Nanyu Li, Li Xu
SegmentationConvolutional Neural NetworkTransformerBiomedical DataComputed Tomography
🎯 What it does: SC-Former was developed for precise segmentation of pulmonary fissures in robotic thoracic surgery.
SCA3D: Enhancing Cross-Modal 3D Retrieval via 3D Shape and Caption Paired Data Augmentation
Junlong Ren, Hao Wang
Data SynthesisRetrievalLarge Language ModelPrompt EngineeringVision Language ModelContrastive LearningTextMultimodalityPoint CloudMesh
🎯 What it does: Proposes the SCA3D method, which uses LLaVA to generate captions for segmented 3D shape components and performs data augmentation to create new 3D-text pairs, thereby enhancing cross-modal 3D retrieval performance.
Scalable Multi-Agent Surveillance: a Kernel-Based Approach
Shashwata Mandal, Sourabh Bhattacharya
OptimizationComputational Efficiency
🎯 What it does: Study the deployment of mobile guard teams to maintain visibility of unpredictable moving intruders, proposing a computationally efficient kernel point generation method to cover polygons, polygon division based on kernel points, control laws for tracking intruders in general polygonal environments, and variations including capture and search, demonstrating team visual coverage improvements through extensive simulations.
Scalable Multi-Robot Task Allocation and Coordination Under Signal Temporal Logic Specifications
Wenliang Liu, Federico Pecora
OptimizationRobotic Intelligence
🎯 What it does: Propose a scalable multi-robot task allocation and coordination algorithm that, under the satisfaction of signal temporal logic (STL) specifications, first generates multiple reference paths using a single-robot motion planner, then computes path allocation and robot progress goals through STL constraints and mixed integer linear programming (MILP), and finally tracks progress using a local controller.
Scalable Multi-Session Visual SLAM in Large-Scale Scenes with Subgraph Optimization
Xiaokun Pan, Guofeng Zhang
Pose EstimationOptimizationSimultaneous Localization and MappingImage
🎯 What it does: Proposed a robust multi-scenario large-scale visual SLAM system that can achieve 6-DoF camera localization and maintain global map consistency over long periods.
Scaling Diffusion Policy in Transformer to 1 Billion Parameters for Robotic Manipulation
Minjie Zhu, Jian Tang
Robotic IntelligenceTransformerDiffusion model
🎯 What it does: Propose the ScaleDP scheme to scale up Diffusion Policy from 10 million to 1 billion parameters within the Transformer architecture, achieving scalability in visual motor control.
SCAM-P: Spatial Channel Attention Module for Panoptic Driving Perception
Gopi Krishna Erabati, Helder Araújo
Object DetectionSegmentationAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: Proposed a SCAM-P multi-task network that can simultaneously perform vehicle detection, drivable area segmentation, and lane segmentation, and enhance feature representation capabilities through a lightweight SCAM module;
Scenario-Based Curriculum Generation for Multi-Agent Driving
Axel Brunnbauer, R. Grosu
Autonomous DrivingReinforcement Learning
🎯 What it does: Created the MATS-Gym framework, which generates variable-number multi-agent traffic scenarios using partial scene specifications and executes them in CARLA. It integrates Scenic and ScenarioRunner with a multi-agent training framework, models interactions using partially observable stochastic games, and combines unsupervised environment design to achieve adaptive auto-curriculum.
Scene-Aware Explainable Multimodal Trajectory Prediction
Pei Liu, Jun Ma
Autonomous DrivingExplainability and InterpretabilityDiffusion modelMultimodality
🎯 What it does: Proposed an explainable conditional diffusion multi-modal trajectory prediction model DMTP to reveal environmental factors and internal mechanisms affecting predictions.
SCU-Hand: Soft Conical Universal Robotic Hand for Scooping Granular Media from Containers of Various Sizes
Tomoya Takahashi, Yoshitaka Ushiku
Robotic Intelligence
🎯 What it does: Designed and verified a soft conical universal manipulator for automatically scooping powder materials from containers of different sizes.
SE-STDGNN: A Self-Evolving Spatial-Temporal Directed Graph Neural Network for Multi-Vehicle Trajectory Prediction
Zixuan Guo, Ben M. Chen
Autonomous DrivingGraph Neural NetworkTime SeriesSequential
🎯 What it does: Proposed a self-evolving spatiotemporal directed graph neural network (SE-STDGNN) for multi-vehicle trajectory prediction
Sea-U-Whale: A Reconfigurable Marine Robot with Multi-Modal Motion
Wendi Ding, Ben M. Chen
Robotic Intelligence
🎯 What it does: Designed and verified a reconfigurable marine robot called Sea-U-Whale that can dynamically adjust its actuator configuration in marine environments, providing three distinct vehicle modes to adapt to various marine scenarios.
SEAL: A Sample-Efficient Adjustment-Learning Method for Table Tennis Robot Serve
Qitong Guo, Yuji Yamakawa
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a sample-efficient adjustment learning (SEAL) method based on human table tennis experience for robot serving tasks, which can enhance training samples without additional sampling. Random interpolation and joint learning in joint space and Cartesian space improve model generalization and robustness, achieving a first-ball positioning error <30mm.
Search-Based Path Planning in Interactive Environments Among Movable Obstacles
Zhongqiang Ren, Ji Zhang
Optimization
🎯 What it does: Developed PAMO*, a search planning method capable of achieving optimal search in environments with movable obstacles, encompassing two planning forms: dual objectives and resource constraints, and proposed a hybrid state version applicable to continuous spaces.
SeaSplat: Representing Underwater Scenes with 3D Gaussian Splatting and a Physically Grounded Image Formation Model
Daniel Yang, Y. Girdhar
GenerationGaussian SplattingImagePhysics Related
🎯 What it does: Propose the SeaSplat method, achieving real-time rendering of underwater scenes using 3D Gaussian Splatting combined with a physics-based underwater image formation model.
Second-Order Stein Variational Dynamic Optimization
Yuichiro Aoyama, Evangelos A. Theodorou
Optimization
🎯 What it does: Proposes a second-order trajectory optimization algorithm based on the Stein Variational Newton method and maximum entropy differential dynamic programming.
Seeing Eye to Eye: Design and Evaluation of a Custom Expressive Eye Display Module for the Stretch Mobile Manipulator
Rafael Morales Mayoral, Naomi T. Fitter
Robotic Intelligence
🎯 What it does: Designed and evaluated a custom expressive LED eye module for a mobile manipulator to demonstrate gaze and emotion expression.
Segment Any Repeated Object
Yushi Liu, Margret Keuper
Object DetectionSegmentationTransformerContrastive LearningImageBenchmark
🎯 What it does: Developed a repeat object segmentation pipeline for open-world scenarios that can detect and assign the same category label to repeated instances of the same object, enabling identification and sorting of similar objects.
Self-Corrective Task Planning by Inverse Prompting with Large Language Models
Jiho Lee, Eunwoo Kim
Explainability and InterpretabilityRobotic IntelligenceTransformerLarge Language ModelPrompt EngineeringChain-of-Thought
🎯 What it does: Propose a self-correcting task planning method that utilizes reverse prompts, combining reasoning steps to generate reverse actions and verify their recoverability to improve the accuracy of robot task planning.
Self-Deformable Magnetic Miniature Robot for Traction Assistance in Endoscopic Submucosal Dissection
Bolan Zhang, Fumihito Arai
Robotic Intelligence
🎯 What it does: Designed and demonstrated a magnetic flexible microrobot that can be safely deployed through an endoscopic channel and provides constant traction to assist in endoscopic submucosal dissection.
Self-Improving Autonomous Underwater Manipulation
Ruoshi Liu, Carl Vondrick
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes AquaBot—a fully autonomous underwater operating system that utilizes behavior cloning and self-learning optimization.
Self-Mixing Laser Interferometry for Robotic Tactile Sensing
R. Proesmans, Francis Wyffels
Robotic IntelligencePhysics Related
🎯 What it does: Designed and verified a robotic fingertip using self-mixing interference (SMI) technology for detecting object sliding and external contact.
Self-Reflective Perceptual Adaptation for Robust Ground Navigation in Unstructured Off-Road Environments
S. Siva, Hao Zhang
Autonomous DrivingRepresentation LearningWorld Model
🎯 What it does: Propose a self-reflective perceptual adaptation method to enhance robust navigation in aerial environments
Self-Sufficient 5-DoF Discrete Global Localization for Magnetically-Actuated Endoscope in Bronchoscopy
Jiewen Tan, Shing Shin Cheng
OptimizationRobotic IntelligenceSimultaneous Localization and MappingImageBiomedical Data
🎯 What it does: Proposes a self-sufficient discrete global localization method based solely on endoscopic images for magnetically actuated endoscopes (MAE)
Self-Supervised Learning of Reconstructing Deformable Linear Objects Under Single-Frame Occluded View
Song Wang, Dan Wu
RestorationRepresentation LearningPoint Cloud
🎯 What it does: Propose a framework combining self-supervised point cloud completion with clustering, sorting, fitting, and deformation of linear objects for occlusion reconstruction, generating ordered keypoints.
Self-Supervised Meta-Learning for All-Layer DNN-Based Adaptive Control with Stability Guarantees
Guanqi He, Guanya Shi
Meta Learning
🎯 What it does: Proposed a full-layer DNN adaptive control framework based on self-supervised meta-learning, which pre-trains the DNN offline through self-supervised meta-learning and updates the entire DNN online to achieve rapid adaptation.
SELP: Generating Safe and Efficient Task Plans for Robot Agents with Large Language Models
Yi Wu, Suresh Jagannathan
Safty and PrivacyRobotic IntelligenceLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposes the SELP framework, which utilizes a large language model (LLM) to generate safe and efficient robot task plans that comply with user natural language commands.
Semantic and Feature Guided Uncertainty Quantification of Visual Localization for Autonomous Vehicles
Qiyuan Wu, Mark Campbell
Autonomous DrivingVideo
🎯 What it does: A lightweight sensor error model based on image features and semantic information was developed for uncertainty quantification in visual localization.
Semantic Cross-Pose Correspondence from a Single Example
Denis Hadjivelichkov, Dimitrios Kanoulas
Pose EstimationImage
🎯 What it does: Propose a method to learn cross-object pose correspondences from a single example, capable of predicting semantically meaningful poses of objects relative to another object; the method achieves this by detecting interacting object parts, performing one-shot part correspondence, and fusing geometric and visual semantic features;
Semantic-Supervised Spatial-Temporal Fusion for LiDAR-Based 3D Object Detection
Chaoqun Wang, Ruimao Zhang
Object DetectionConvolutional Neural NetworkSupervised Fine-TuningPoint Cloud
🎯 What it does: Proposed a semantic-supervised spatiotemporal fusion method called ST-Fusion to enhance LiDAR-based 3D object detection.
Semi-Autonomous 2.5D Control of Untethered Magnetic Suture Needle
Qinhan Wang, Axel Krieger
Object DetectionPose EstimationRobotic IntelligenceBiomedical Data
🎯 What it does: This paper develops and implements a semi-autonomous 2.5D control method for a cordless magnetic suture needle, utilizing an external electromagnetic manipulator to control the needle tip's planar position and in/out-of-plane orientation, completing suturing tasks on a simulated tissue gel.
Semi-Elastic LiDAR-Inertial Odometry
Zikang Yuan, Xin Yang
Autonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed and implemented a semi-elastic optimized LiDAR-IMU state estimation method, integrating it into a self-developed optimized LiDAR-IMU odometry framework.
Sensory Glove-Based Surgical Robot User Interface
L. Borgioli, P. Giulianotti
Robotic IntelligenceMultimodality
🎯 What it does: Propose a system that integrates the HTC Vive tracker, Manus Meta Prime 3 XR sensing gloves, and SCOPEYE wireless smart glasses, using hand movements and gestures to intuitively control the single arm and end-effector of the da Vinci surgical robot, supporting tool orientation locking and unlocking, and confirming commands through tactile feedback from the gloves.
Shadow Program Inversion with Differentiable Planning: A Framework for Unified Robot Program Parameter and Trajectory Optimization
Benjamin Alt, Michael Beetz
OptimizationRobotic Intelligence
🎯 What it does: Proposed a first-order optimizer called SPI-DP, which integrates differentiable Gaussian process motion planning (dGPMP2-ND) to jointly optimize the parameters and trajectories of robot programs, satisfying high-level task objectives and motion-level constraints.
ShadowTac: Dense Measurement of Shear and Normal Deformation of a Tactile Membrane from Colored Shadows
Giuseppe Vitrani, Michaël Wiertlewski
Robotic IntelligenceOptical FlowImage
🎯 What it does: Designed and evaluated a novel tactile sensor called ShadowTac, capable of simultaneously densely measuring the normal and shear deformations of the touch membrane, and estimating the initial sliding of any object through these measurements;
Shape-Programming Robotic Reflectors for Wireless Networks
Yawen Liu, Swarun Kumar
Robotic Intelligence
🎯 What it does: Designed and experimentally validated a shape-programmable robotic reflector (MetaMorph) to enhance signal quality in low-power wide-area networks (LP-WAN).
Shape-Space Deformer: Unified Visuo-Tactile Representations for Robotic Manipulation of Deformable Objects
Sean M. V. Collins, Peyman Moghadam
Representation LearningRobotic IntelligenceMultimodality
🎯 What it does: Proposes Shape-Space Deformer, which uses template augmentation to construct a unified visuo-tactile representation, enabling fine-grained and robust reconstruction of various deformations, maintaining strong generalization capabilities under unseen force conditions, and rapidly adapting to new objects.
Shared Control for Cable Routing with Tactile Sensing
Ange Bao, Pei Zhao
Robotic Intelligence
🎯 What it does: Proposed a shared control method based on tactile perception for multi-stage, contact-rich cable laying tasks, validated in tasks such as straightening cables and inserting them into clamps.
SHIRE: Enhancing Sample Efficiency using Human Intuition in REinforcement Learning
Amogh Joshi, Kaushik Roy
Explainability and InterpretabilityReinforcement Learning
🎯 What it does: Propose the SHIRE framework, which encodes human intuition as a probabilistic graphical model (PGM) and embeds it into the deep reinforcement learning training process to improve sample efficiency and interpretability.
SIGMA: Sheaf-Informed Geometric Multi-Agent Pathfinding
Shuhao Liao, G. Sartoretti
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a decentralized deep reinforcement learning framework based on Sheaf theory for solving multi-agent pathfinding (MAPF) problems, finding the shortest and collision-free paths in known environments with obstacles.
Sim2real Within 5 Minutes: Efficient Domain Transfer with Stylized Gaussian Splatting for Endoscopic Images
Junyang Wu, Guangyao Yang
Image TranslationDomain AdaptationGaussian SplattingImageBiomedical DataComputed Tomography
🎯 What it does: Efficient domain transfer from pre-surgical CT domain to intraoperative endoscopic image domain is achieved through Stylized Gaussian Splatting, requiring only a small number of real images and enabling fast training.
Sim4EndoR: A Reinforcement Learning Centered Simulation Platform for Task Automation of Endovascular Robotics
Tianliang Yao, P. Qi
Robotic IntelligenceReinforcement LearningBiomedical Data
🎯 What it does: Developed a reinforcement learning-based endovascular surgical robot simulation platform named Sim4EndoR for the automation of PCI tasks.
SIMP: Real-Time Energy and Time-Efficient 3D Motion Planning for Bio-Inspired AUVs
August Sletnes Bjørlo, Eleni Kelasidi
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Proposed a real-time motion planning framework called SIMP, capable of generating energy- and time-efficient paths with empirical local optimality for articulated swimming robots in three-dimensional space.
Simplifying Reward Design in Complex Robotics: Average-Reward Maximum Entropy Reinforcement Learning
Jean Seong Bjorn Choe, Jong-Kook Kim
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a novel AR-EAPO algorithm for addressing the oscillation ascent and stabilization tasks in underactuated double pendulum systems.
Simultaneous Collision Detection and Force Estimation for Dynamic Quadrupedal Locomotion
Ziyi Zhou, K. Berntorp
Robotic Intelligence
🎯 What it does: This paper addresses the problem of collision detection and force estimation for quadruped robots under dynamic gaits, utilizing joint encoder information and robot dynamics for estimation, and designs corresponding reflexive movements and admittance controllers.
Simultaneous Ground Reaction Force and State Estimation Via Constrained Moving Horizon Estimation
Ji-Min Kang, Xiaobin Xiong
OptimizationRobotic Intelligence
🎯 What it does: Proposes a framework for simultaneously estimating ground reaction forces (GRF) and robot states, employing Constrained Moving Horizon Estimation (MHE) to integrate robot dynamics, sensor data, and contact complementary constraints within a convex window optimization, thereby providing accurate GRF and state estimation.
Simultaneous Localization and Affordance Prediction of Tasks from Egocentric Video
Zachary Chavis, Stephen J. Guy
Robotic IntelligenceVision Language ModelVideo
🎯 What it does: Proposes a method to extend audio-visual language models (VLM) to the spatial dimension, utilizing spatially localized first-person video demonstrations to predict the spatial applicability of tasks and their localization relative to the observer.
Single-Fiber Optical Frequency Domain Reflectometry (Ofdr) Shape Sensing of Continuum Manipulators With Planar Bending
Mobina Tavangarifard, F. Alambeigi
Pose EstimationRobotic Intelligence
🎯 What it does: Proposed a shape sensing device using a single OFDR fiber attached to a flat titanium alloy wire, integrated into a 170mm long flexible continuous mechanical hand, and conducted C, J, S shape reconstruction experiments.
Single-Shot Metric Depth from Focused Plenoptic Cameras
Blanca Lasheras-Hernandez, Javier Civera
Depth EstimationImagePoint Cloud
🎯 What it does: Proposed a single-image dense metric depth estimation pipeline based on a focused plenoptic camera
Single-Stage Optimization of Open-Loop Stable Limit Cycles with Smooth, Symbolic Derivatives
Muhammad Saud Ul Hassan, Christian Hubicki
OptimizationRobotic Intelligence
🎯 What it does: Proposes a general framework that utilizes single-stage constrained optimization with Direct Collocation to rapidly generate open-loop stable limit cycles, capable of enforcing arbitrary compact stability constraints.
Sketch-MoMa: Teleoperation for Mobile Manipulator via Interpretation of Hand-Drawn Sketches
Kosei Tanada, Takashi Yamamoto
Robotic IntelligenceVision Language ModelImageText
🎯 What it does: Proposed a system called Sketch-MoMa for teleoperating a mobile manipulator using hand-drawn sketches.
Skills Made to Order: Efficient Acquisition of Robot Cooking Skills Guided by Multiple Forms of Internet Data
Mrinal Verghese, Christopher G. Atkeson
Data-Centric LearningRobotic IntelligenceLarge Language ModelOptical FlowVideoTextMultimodality
🎯 What it does: Explored the use of internet data sources and foundational models for template behavior selection in robot cooking skills, investigated three template selection methods, and combined them into an efficient template selector, ultimately achieving a 79% success rate in 16 cooking skills involving tool usage.
SKOOTR: A Skating, Omni-Oriented, Tripedal Robot
Adam Hung, Talia Y. Moore
Robotic Intelligence
🎯 What it does: Designed and implemented a three-legged robot named SKOOTR with friction and rolling contact, featuring a freely rotatable central sphere, capable of maintaining stable tri-pod support and achieving various forward gaits, turning, obstacle crossing, and staircase climbing;
SLABIM: A SLAM-BIM Coupled Dataset in HKUST Main Building
Haoming Huang, Huan Yin
Simultaneous Localization and MappingMultimodalityBenchmark
🎯 What it does: This paper designs and constructs the first dataset combining SLAM and BIM, named SLABIM, collects multi-sensor and multi-scenario data and builds models at the main building of the Hong Kong University of Science and Technology, followed by experimental verification on three tasks: registration, localization, and semantic mapping.
SLAM Assisted 3D Tracking System for Laparoscopic Surgery
Jingwei Song, Maani Ghaffari
Pose EstimationSimultaneous Localization and MappingBiomedical Data
🎯 What it does: Propose a real-time monocular 3D tracking algorithm based on ORB-SLAM2, incorporating geometric priors. A pseudo-segmentation strategy is used for rapid initialization and separation of the target organ, followed by integrating the original 3D shape as a geometric prior in pose graph optimization.
Slamspoof: Practical Lidar Spoofing Attacks on Localization Systems Guided by Scan Matching Vulnerability Analysis
Rokuto Nagata, Kentaro Yoshioka
Autonomous DrivingAdversarial AttackSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Designed and implemented SLAMSpoof, demonstrating LiDAR spoofing attacks against autonomous driving positioning systems, and evaluated their impact on positioning accuracy.
SliceOcc: Indoor 3D Semantic Occupancy Prediction with Vertical Slice Representation
Jianing Li, Shanghang Zhang
SegmentationRepresentation LearningConvolutional Neural NetworkTransformerImage
🎯 What it does: Propose the longitudinal slice representation and the SliceOcc model to achieve indoor 3D semantic occupancy prediction using RGB cameras.
SMART: Advancing Scalable Map Priors for Driving Topology Reasoning
Junjie Ye, Liu Ren
Autonomous DrivingSupervised Fine-TuningImageBenchmark
🎯 What it does: Proposes SMART, which leverages Standard Definitions (SD) and satellite maps to learn a map prior model for scalable driving topology reasoning.
Social Gesture Recognition in spHRI: Leveraging Fabric-Based Tactile Sensing on Humanoid Robots
Dakarai Crowder, Wenzhen Yuan
RecognitionRobotic IntelligenceTime Series
🎯 What it does: Developed a social gesture recognition system based on fabric-based large-scale tactile sensors, and constructed a social gesture dataset with multiple participants
Social-MAE: Social Masked Autoencoder for Multi-Person Motion Representation Learning
Mahsa Ehsanpour, Hamid Rezatofighi
Pose EstimationRepresentation LearningTransformerSupervised Fine-TuningAuto EncoderSequential
🎯 What it does: Propose the Social-MAE Transformer, a pre-training framework based on masked autoencoders for multi-person human motion representation learning, which is fine-tuned on multiple high-level social tasks.
Socratic Planner: Self-QA-Based Zero-Shot Planning for Embodied Instruction Following
Suyeon Shin, Byoung-Tak Zhang
Robotic IntelligenceLarge Language ModelVision-Language-Action ModelTextMultimodalityBenchmark
🎯 What it does: Propose Socratic Planner, a zero-shot planning method based on self-asking and self-answering for executing natural language instructions
SOF-E: An Energy Efficient Robot for Collaborative Transport and Placement of Mechanical Meta-Material Modules
Inchul Moon, Kenneth C. Cheung
Computational EfficiencyRobotic Intelligence
🎯 What it does: Proposed and experimentally tested a five-degree-of-freedom, low-mass robot called SOF-E for collaborative transportation and placement of mechanical metamaterial modules, demonstrating its potential applications in space assembly.
Soft Actor-Critic-Based Control Barrier Adaptation for Robust Autonomous Navigation in Unknown Environments
Nicholas Mohammad, N. Bezzo
Autonomous DrivingReinforcement Learning
🎯 What it does: Propose a strategy based on Soft Actor-Critic (SAC) that dynamically adjusts control barrier function (CBF) constraint parameters during runtime to achieve safe and non-conservative motion planning.
Soft Robotic Dynamic in-Hand Pen Spinning
Yunchao Yao, Jeffrey Ichnowski
Robotic Intelligence
🎯 What it does: Learning soft-hand grasping and rotation primitives through self-labeled experiments in the real world to achieve pen rotation.
SOLVR: Submap Oriented LiDAR-Visual Re-Localisation
Joshua Knights, Stefan Leutenegger
Pose EstimationDepth EstimationRetrievalAutonomous DrivingSimultaneous Localization and MappingImageMultimodalityPoint Cloud
🎯 What it does: Introduces the SOLVR integrated pipeline for learning-based LiDAR-visual relocalization, achieving cross-modal place recognition and 6-DoF registration.
Space-Aware Instruction Tuning: Dataset and Benchmark for Guide Dog Robots Assisting the Visually Impaired
ByungOk Han, Jaehong Kim
Robotic IntelligenceSupervised Fine-TuningVision Language ModelMultimodalityBenchmark
🎯 What it does: Proposed the Space-Aware Instruction Tuning (SAIT) dataset and the Space-Aware Benchmark (SA-Bench), and performed spatial-aware instruction tuning on Vision-Language Models (VLMs) to improve navigation guidance for visually impaired individuals by blind navigation robots.
SparseDrive: End-to-End Autonomous Driving via Sparse Scene Representation
Wenchao Sun, Sifa Zheng
Autonomous Driving
🎯 What it does: Proposed SparseDrive end-to-end autonomous driving system, which includes symmetric sparse perception module and parallel motion planner, unifying detection, tracking, online mapping, and motion prediction and planning
Spatial Sensitivity Equalization of ERT-Based Robotic Skin Through Gauge Factor Distribution Optimization
Junhwi Cho, Jung Kim
OptimizationRobotic Intelligence
🎯 What it does: Regulating the conductivity of ERT sensors through topology optimization to achieve spatial sensitivity balance, and fabricating complex conductive patches using screen printing technology; validating the effectiveness of this method compared to traditional ERT sensors in both simulation and real-world environments.
SpatialBot: Precise Spatial Understanding with Vision Language Models
Wenxiao Cai, Bo Zhao
Vision Language ModelMultimodalityBenchmark
🎯 What it does: Designed and trained the SpatialBot model, leveraging RGB and depth images to enhance the spatial understanding capabilities of Vision-Language Models (VLMs).
Spatially Constrained and Deeply Learned Bilateral Structural Intensity-Depth Registration Autonomously Navigates a Flexible Endoscope
Hao Fang, Xióngbiao Luó
Depth EstimationOptimizationRobotic IntelligenceConvolutional Neural NetworkImageBiomedical Data
🎯 What it does: Proposed and implemented a spatial-constrained, deep learning-driven dual-structure intensity-depth 2D-3D registration framework for autonomous navigation of flexible endoscopes.
Speedtuning: Speeding Up Policy Execution with Lightweight Reinforcement Learning
David Yuan, Chelsea Finn
Robotic IntelligenceReinforcement Learning
🎯 What it does: Developed the Speed Tuning framework to enhance the execution speed of robotic manipulation strategies
SPIBOT: A Drone-Tethered Mobile Gripper for Robust Aerial Object Retrieval in Dynamic Environments
Gyuree Kang, D. Shim
Robotic Intelligence
🎯 What it does: Designed a drone tethered mobile gripper SPIBOT for achieving robust target grasping in dynamic environments.
SPINE: Online Semantic Planning for Missions with Incomplete Natural Language Specifications in Unstructured Environments
Zachary Ravichandran, Vijay Kumar
OptimizationRobotic IntelligenceTransformerLarge Language ModelText
🎯 What it does: Proposes SPINE, a semantic planning framework capable of online processing of incomplete natural language task descriptions, leveraging large language models (LLMs) for subtask reasoning within a rolling horizon framework, while automatically verifying safety and online refining after acquiring new map observations.
SplatSim: Zero-Shot Sim2Real Transfer of RGB Manipulation Policies Using Gaussian Splatting
M. N. Qureshi, Abhisesh Silwal
Domain AdaptationGaussian SplattingImage
🎯 What it does: Propose the SplatSim framework, which uses Gaussian splat rendering instead of traditional meshes to generate highly realistic and scalable RGB composite images, and deploy RGB-based manipulation strategies in real environments without prior training (zero-shot).
SPOT: SE(3) Pose Trajectory Diffusion for Object-Centric Manipulation
Cheng-Chun Hsu, S. Birchfield
Robotic IntelligenceDiffusion model
🎯 What it does: Propose the SPOT framework, which uses object-centric SE(3) pose trajectories as object-centric task representations, training diffusion policies to accomplish object manipulation tasks.
SR-AIF: Solving Sparse-Reward Robotic Tasks From Pixels with Active Inference and World Models
Viet Dung Nguyen, Alexander Ororbia
Robotic IntelligenceReinforcement LearningWorld ModelImage
🎯 What it does: For sparse reward, continuous action, goal-based robot control POMDP, the active inference (AIF) method is used, incorporating prior preference learning technology and self-revision planning.
SRL-Gym: A Morphology and Controller Co-Optimization Framework for Supernumerary Robotic Limbs in Load-Bearing Locomotion
Lingyi Meng, Zhong Zhang
OptimizationRobotic Intelligence
🎯 What it does: Propose and verify a two-layer optimization-based morphology-controller co-optimization framework for automatically generating and optimizing the structure of super-large robot arms (SRL) to adapt to load-bearing gait tasks
SSF: Sparse Long-Range Scene Flow for Autonomous Driving
Ajinkya Khoche, P. Jensfelt
Autonomous DrivingConvolutional Neural NetworkPoint Cloud
🎯 What it does: Propose the Sparse Scene Flow (SSF) approach, utilizing a sparse convolution backbone to achieve long-range scene flow estimation.
Stable Tracking of Eye Gaze Direction During Ophthalmic Surgery
Tinghe Hong, Kai Huang
Pose EstimationRobotic IntelligenceBiomedical Data
🎯 What it does: Propose a method combining machine learning with traditional algorithms for eye localization and tracking, eliminating the dependency on feature points, enabling stable iris detection and estimation of eye direction under varying lighting and shadow conditions, used to guide surgical robot control.
Stage-Wise Reward Shaping for Acrobatic Robots: A Constrained Multi-Objective Reinforcement Learning Approach
Dohyeong Kim (Seoul National University), Songhwai Oh (Seoul National University)
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a reward and cost function divided into stages, shaping rewards and controlling acrobatic robots under a constrained multi-objective reinforcement learning framework, implementing a corresponding practical algorithm, and verifying its effectiveness in both simulated and real environments.
Stair Climbing of a Transformable Robot Using Varying Leg-Wheel Contact Points
Yen-Li Lai, Pei-Chun Lin
Robotic Intelligence
🎯 What it does: Developed a behavior strategy for leg-wheel reconfigurable robots to climb stairs.
Stands on Shoulders of Giants: Learning to Lift 2D Detection to 3D with Geometry-Driven Objectives
Jhih-Rong Chen, Wei-Chen Chiu
Object DetectionConvolutional Neural NetworkImage
🎯 What it does: Propose a method that lifts 2D detection results into 3D space to achieve 3D vehicle detection;
STEER: Flexible Robotic Manipulation via Dense Language Grounding
Laura Smith, Ted Xiao
Robotic IntelligenceLarge Language ModelVision Language ModelVision-Language-Action ModelTextMultimodality
🎯 What it does: Proposed the STEER framework, which integrates high-level common sense reasoning with low-level fine-grained control through a densely annotated language benchmark strategy, enabling robots to achieve intelligent adaptation in complex scenarios.
Steerable Tape-Spring Needle for Autonomous Sharp Turns Through Tissue
Omar T. Abdoun, Mark Yim
Robotic IntelligenceBiomedical DataUltrasound
🎯 What it does: Proposed a new controllable tape-spring needle capable of achieving sharp turns between 15 to 150 degrees with a minimum turning radius of 3 mm, derived and experimentally validated its geometric model; evaluated manual and robotic Dubins path navigation of the needle in 7 kPa and 13 kPa tissue models to investigate its robustness in non-uniform tissue hardness, and demonstrated needle tip tracking via ultrasound imaging, anatomical obstacle avoidance, and integration with robotic autonomous navigation systems.
SteeredMarigold: Steering Diffusion Towards Depth Completion of Largely Incomplete Depth Maps
Jakub Gregorek, L. Nalpantidis
Depth EstimationDiffusion modelImage
🎯 What it does: Proposes SteeredMarigold, a training-free, zero-shot depth completion method that can generate metrically dense depth maps for large-scale missing depth areas.
Steering Prediction via a Multi-Sensor System for Autonomous Racing
Zhuyun Zhou, Tobi Delbruck
Autonomous DrivingMultimodalityPoint Cloud
🎯 What it does: Integrate 2D LiDAR and event camera data within an end-to-end learning framework for steering prediction, create a multi-sensor dataset, benchmark SOTA fusion methods, and propose a low-rank efficient fusion design along with a novel fusion learning strategy.
Stiffness Regulation Co-Pilot in Bilateral Teleimpedance Control: A Preliminary User Study
Pedro Gomez Hernandez, Cheng Fang
Robotic Intelligence
🎯 What it does: Studied the concept of a collaborative driver with variable stiffness regulation in bidirectional remote impedance control, and evaluated its effectiveness through a pre-experiment.
Stochastic Trajectory Prediction Under Unstructured Constraints
Hao Ma, Jianqiang Yi
GenerationAutonomous DrivingDiffusion modelScore-based ModelTime SeriesSequentialBenchmark
🎯 What it does: A method for trajectory prediction under unstructured constraints was studied, i.e., achieving constrained trajectory prediction using conditional diffusion models.
Stonefish: Supporting Machine Learning Research in Marine Robotics
Michele Grimaldi, Nuno Gracias
Robotic Intelligence
🎯 What it does: Multiple new sensors (event camera, thermal camera, optical flow camera) as well as visual optical communication, towed equipment support, improved thruster modeling, more flexible fluid dynamics models, and more accurate sonar simulations were added to the open-source simulation platform Stonefish, along with an automatic annotation tool.
Stop-N-Go: Search-Based Conflict Resolution for Motion Planning of Multiple Robotic Manipulators
Gidon Han, Changjoo Nam
OptimizationRobotic Intelligence
🎯 What it does: Propose a conflict resolution method that inserts pauses into pre-planned multi-robot trajectories using A* search to minimize total completion time.
Strain-Coordinated Formation, Migration, and Encapsulation Behaviors in a Tethered Robot Collective
Sadie Cutler, Kirstin Petersen
Robotic Intelligence
🎯 What it does: Explore the use of flexible, passive, fixed-length connections as sensors to achieve distributed morphological control, extending to confinement and migration along global gradients, and study the trade-off between morphological control and approach behavior in obstacle environments.
Strategic System Design for High Precision in Assembly Processes of CPU
Cheuk Tung Shadow Yiu, Kam Tim Woo
Pose EstimationRobotic Intelligence
🎯 What it does: Designed and implemented a high-precision robotic system for CPU assembly, focusing on device selection and sensor parameter optimization, and integrating geometric segmentation, binocular structured light imaging, and 6D pose estimation based on 3D information.
Stretchable Electrohydraulic Artificial Muscle for Full Motion Ranges in Musculoskeletal Antagonistic Joints
Amirhossein Kazemipour, Robert K. Katzschmann
🎯 What it does: Proposed a contractile and extensible adversarial muscle system, combining non-extensible electrohydraulic soft actuators with electrostatic clutches to support full range of motion.
Structure-Aware Radar-Camera Depth Estimation
Fuyi Zhang, Hui-liang Shen
Depth EstimationAutonomous DrivingImageMultimodalityPoint Cloud
🎯 What it does: Proposed a structure-based radar depth enhancement strategy and a multi-scale structure-guided network, and constructed the structure-aware radar-camera depth estimation framework SA-RCD
Subassembly to Full Assembly: Effective Assembly Sequence Planning Through Graph-Based Reinforcement Learning
Chang Shu, Shinkyu Park
Robotic IntelligenceGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: Propose a framework named Subassembly to Assembly (S2A), enabling robots to assemble multi-component objects through object manipulation actions within a predefined structure.
Submodular Optimization for Keyframe Selection & Usage in SLAM
David Thorne, B. T. Lopez
OptimizationSimultaneous Localization and Mapping
🎯 What it does: Proposes two novel keyframe selection strategies and a new submap generation method, providing a fast map summarization functionality for capturing the environment.
SuFIA-BC: Generating High Quality Demonstration Data for Visuomotor Policy Learning in Surgical Subtasks
M. Moghani, Animesh Garg
Data SynthesisRobotic IntelligenceImageVideoBiomedical Data
🎯 What it does: Proposes the SuFIA-BC framework, which generates high-quality synthetic data in a comprehensive simulator using enhanced surgical digital twins and optically realistic human organs to address surgical autonomous tasks; simultaneously investigates 3D visual representations of visual observation spaces derived from multi-view cameras and single-endoscopic camera views, and validates the challenges of these tasks for behavior cloning models through systematic evaluation.
Suite-IN: Aggregating Motion Features from Apple Suite for Robust Inertial Navigation
Lan Sun, Ling Pei
Autonomous DrivingContrastive LearningTime Series
🎯 What it does: Proposed a multi-device deep learning framework called Suite-IN, which aggregates motion data from Apple Suite for inertial navigation;
SuperLoc: The Key to Robust Lidar-Inertial Localization Lies in Predicting Alignment Risks Superodometry.Com/SuperLoc
Shibo Zhao, Sebastian A. Scherer
Autonomous DrivingSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposes SuperLoc, aimed at enhancing the robustness of map-based LiDAR localization in degraded environments.
SuperQ-GRASP: Superquadrics-Based Grasp Pose Estimation on Larger Objects for Mobile-Manipulation
Xun Tu, K. Desingh
Pose EstimationRobotic IntelligenceNeural Radiance FieldImageMesh
🎯 What it does: Proposes a grasp pose estimation workflow based on superquadrics, which utilizes NeRF to reconstruct an implicit model from RGB images, extracts explicit meshes, and decomposes them into superquadrics, thereby combining precomputed grasp poses with geometric primitives corresponding to the target object.