arXivSub Start free trial

ICRA 2024 Papers — Page 15

IEEE International Conference on Robotics and Automation · 1760 papers

Salience-guided Ground Factor for Robust Localization of Delivery Robots in Complex Urban Environments

Jooyong Park, Younggun Cho

Autonomous DrivingSimultaneous Localization and Mapping

🎯 What it does: Propose a method that utilizes significant object detection (SOD) to extract unique features in urban environments, and achieves consistent ground features through motion compensation inverse perspective mapping (MC-IPM) and moment calculation, to enhance the performance of delivery robots in loop closure and localization.

SAM-Event-Adapter: Adapting Segment Anything Model for Event-RGB Semantic Segmentation

Bowen Yao, Zheng Yang

SegmentationTransformerMultimodality

🎯 What it does: By introducing a lightweight SAM-Event-Adapter module, event camera data is integrated into the cross-modal learning architecture of the Segment Anything Model (SAM), achieving event-RGB semantic segmentation.

Sample-Efficient Learning to Solve a Real-World Labyrinth Game Using Data-Augmented Model-Based Reinforcement Learning

Thomas Bi, Raffaello D'Andrea

OptimizationRobotic IntelligenceReinforcement LearningWorld ModelImage

🎯 What it does: Developed and trained a model-based reinforcement learning (RL) robot system that uses low-dimensional observations to extract and crop image patches, guiding the robot to navigate and solve maze games in a physical maze. The system leverages maze progress as a reward and employs data augmentation by leveraging system symmetry, ultimately solving the maze in record time using only 5 hours of real training data.

SANet: Small but Accurate Detector for Aerial Flying Object

Xunkuai Zhou, Ben M. Chen

Object DetectionConvolutional Neural NetworkImage

🎯 What it does: Proposed SANet, a detector for small-sized, high-precision detection of aerial objects.

SARA-RT: Scaling up Robotics Transformers with Self-Adaptive Robust Attention

Isabel Leal, Kanishka Rao

Robotic IntelligenceTransformerSupervised Fine-TuningVision-Language-Action ModelMultimodality

🎯 What it does: proposes the SARA-RT framework, which utilizes the up-training method to convert Transformer-based robot policies (including large-scale vision-language-action models) with originally quadratic time complexity into a high-quality linear attention version, thereby enabling efficient deployment of robot systems.

SATac: A Thermoluminescence Enabled Tactile Sensor for Concurrent Perception of Temperature, Pressure, and Shear

Ziwu Song, Wenbo Ding

Robotic IntelligenceMultimodalityPhysics Related

🎯 What it does: Designed a multi-modal visual tactile sensor named SATac, capable of simultaneously perceiving temperature, pressure, and shear;

Saturation-Aware Angular Velocity Estimation: Extending the Robustness of SLAM to Aggressive Motions*

Simon-Pierre Deschênes, François Pomerleau

Autonomous DrivingRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed a method using accelerometers to estimate angular velocity, enhancing the robustness of SLAM in scenarios where intense motion causes gyroscope saturation

Scalable Multi-Robot Collaboration with Large Language Models: Centralized or Decentralized Systems?

Yongchao Chen, Chuchu Fan

Robotic IntelligenceTransformerLarge Language ModelAgentic AIPrompt Engineering

🎯 What it does: Using a pre-trained large language model (LLM) as a planner for multi-robot collaborative tasks, and comparing the performance of different multi-agent communication frameworks in multi-task scenarios

Scalable underwater assembly with reconfigurable visual fiducials

Samuel E. Lensgraf, Alberto Quattrini Li

Robotic IntelligenceImage

🎯 What it does: Propose a scalable underwater assembly system that autonomously adjusts its visual marker infrastructure to adapt to different manipulation tasks and uses AUVs to complete assembly operations.

SCALE: Self-Correcting Visual Navigation for Mobile Robots via Anti-Novelty Estimation

Chang Chen, Shunbo Zhou

Autonomous DrivingRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a self-correcting visual navigation method called SCALE, which can proactively avoid out-of-distribution (OOD) scenarios without human intervention.

Scaling Motion Forecasting Models with Ensemble Distillation

Scott Ettinger, Rami Al-Rfou

Autonomous DrivingKnowledge DistillationTime SeriesSequential

🎯 What it does: Propose to enhance motion prediction systems under computational budget constraints through model ensemble and distillation techniques

Scaling Object-centric Robotic Manipulation with Multimodal Object Identification

Chaitanya Mitash, Kapil D. Katyal

RecognitionRobotic IntelligenceVision Language ModelContrastive LearningMultimodalityBenchmark

🎯 What it does: Propose using a multimodal (image and text) reference database for object recognition, and design a training strategy that learns domain-invariant image embeddings, image-text matching, and multi-source prediction fusion.

Scaling Team Coordination on Graphs with Reinforcement Learning

Manshi Limbu, Xuesu Xiao

Reinforcement LearningGraph

🎯 What it does: This paper converts team coordination and support action problems in a graph environment into a Markov Decision Process (MDP), enabling multi-agent systems to learn support behaviors and minimum-cost path planning on graphs through reinforcement learning (RL).

Scene Action Maps: Behavioural Maps for Navigation without Metric Information

Joel Loo, D. Hsu

Robotic IntelligenceWorld ModelImage

🎯 What it does: Propose Scene Action Map (SAM), a behavior-based topological graph, and develop a learnable map reading method to convert 2D maps into SAM; build and deploy a behavior navigation stack on a quadruped robot, and evaluate its navigation performance.

Scene Informer: Anchor-based Occlusion Inference and Trajectory Prediction in Partially Observable Environments

Bernard Lange, Mykel J. Kochenderfer

Autonomous DrivingTransformerMultimodality

🎯 What it does: Propose Scene Informer, a unified framework for predicting the trajectories of visible agents and inferring occluded agents in partially observable environments.

SceneControl: Diffusion for Controllable Traffic Scene Generation

Jackie Lu, R. Urtasun

GenerationAutonomous DrivingDiffusion modelImage

🎯 What it does: Proposed the SceneControl framework for controllable traffic scene generation;

SceneReplica: Benchmarking Real-World Robot Manipulation by Creating Replicable Scenes

Ninad Khargonkar, Yu Xiang

Robotic IntelligenceBenchmark

🎯 What it does: Proposed a reproducible real-world robotic manipulation benchmark focused on grasping and placing tasks.

SculptBot: Pre-Trained Models for 3D Deformable Object Manipulation

Alison Bartsch, A. Farimani

Computational EfficiencyRepresentation LearningRobotic IntelligenceTransformerWorld ModelPoint Cloud

🎯 What it does: This paper addresses the challenges of soft object manipulation by proposing a point cloud-based state representation and utilizing a pre-trained point cloud reconstruction transformer to learn a latent dynamics model for predicting material deformation under grasping actions; it also designs an action sampling algorithm based on point cloud geometric differences to enhance model-based planning efficiency and completes sculpture experiments with various simple shapes in real environments.

SE(2) Assembly Planning for Magnetic Modular Cubes

K. Keune, Aaron Becker

Robotic IntelligenceGraphPhysics Related

🎯 What it does: For magnetic modular blocks with embedded permanent magnets, this paper designs local planning and closed-loop control algorithms to achieve the connection of two structures on a specified surface, and proposes a global planner to generate a construction instruction graph. Subsequently, the target structure is built by repeatedly applying local planning through depth-first search; meanwhile, the impact of structural size and shape on planning time is analyzed.

Sea-U-Foil: A Hydrofoil Marine Vehicle with Multi-Modal Locomotion

Zuoquan Zhao, Ben M. Chen

Robotic Intelligence

🎯 What it does: Designed a remote-controlled hydrofoil marine vehicle named Sea-U-Foil, which has three motion modes: displacement mode, wing-load mode, and submersion mode.

Seabed intervention with an underwater legged robot

G. Picardi, M. Calisti

Robotic Intelligence

🎯 What it does: Conducted an on-site demonstration of remote operation grasping and placing tasks for the underwater legged robot SILVER2.

See to Touch: Learning Tactile Dexterity through Visual Incentives

Irmak Guzey, Lerrel Pinto

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningContrastive LearningImage

🎯 What it does: Propose the TAVI framework, which optimizes tactile-based multi-finger robot grasping and manipulation strategies using visual rewards

Self Model for Embodied Intelligence: Modeling Full-Body Human Musculoskeletal System and Locomotion Control with Hierarchical Low-Dimensional Representation

Kaibo He, Yanan Sui

Robotic IntelligenceReinforcement LearningBiomedical Data

🎯 What it does: Constructed a complete human musculoskeletal model (MS-HUMAN-700) with 90 body segments, 206 joints, and 700 tendon units, and developed a whole-body control algorithm using low-dimensional representation and hierarchical deep reinforcement learning.

Self-Reconfigurable Robots for Collaborative Discrete Lattice Assembly

Miana Smith, N. Gershenfeld

Robotic Intelligence

🎯 What it does: A self-reconfigurable robotic system was developed for collaborative assembly of 3D discrete lattice structures, supporting self-replication of the robots themselves.

Self-Recovery Prompting: Promptable General Purpose Service Robot System with Foundation Models and Self-Recovery

Mimo Shirasaka, Yusuke Iwasawa

Robotic IntelligencePrompt EngineeringMultimodality

🎯 What it does: Developed a general-purpose service robot system for the global competition RoboCup@Home2023, and proposed a self-recovery prompting process;

Self-Righting Shell for Robotic Hexapod

Katelyn King, Shai Revzen

Robotic Intelligence

🎯 What it does: Designed and implemented a geometric shell that enables the six-legged robot MediumANT to passively self-right, and fixed it to the robot chassis

Self-Sensing Feedback Control of an Electrohydraulic Robotic Shoulder

Clemens C. Christoph, Robert K. Katzschmann

Robotic Intelligence

🎯 What it does: Developed a self-sensing electro-hydraulic actuated bio-inspired shoulder joint with two degrees of freedom capable of achieving closed-loop trajectory control.

Self-supervised 6-DoF Robot Grasping by Demonstration via Augmented Reality Teleoperation System

Xiwen Dengxiong, Yunbo Zhang

Pose EstimationRobotic IntelligenceContrastive Learning

🎯 What it does: Propose a self-supervised 6-degree-of-freedom (6-DoF) grasping pose detection framework based on augmented reality (AR) remote operation, which can learn grasping strategies from human demonstrations and provide grasping poses without requiring annotated grasping pose labels.

Self-supervised Learning for Joint Pushing and Grasping Policies in Highly Cluttered Environments

Yongliang Wang, H. Kasaei

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a deep reinforcement learning method that learns a joint grasp and push strategy to effectively manipulate target objects in highly crowded environments.

Self-Supervised Learning of Monocular Visual Odometry and Depth with Uncertainty-Aware Scale Consistency

Changhao Wang, Wei Zhou

Pose EstimationDepth EstimationVideo

🎯 What it does: Proposes a feature-based visual odometry learning system with an effective scale recovery strategy, and introduces a photometric sensitivity depth uncertainty learning method for guiding scale recovery. Evaluation shows that it can predict scale-consistent trajectories from monocular videos and achieve state-of-the-art performance on KITTI.

Self-supervised Pretraining and Finetuning for Monocular Depth and Visual Odometry

Boris Chidlovskii, Leonid Antsfeld

Pose EstimationDepth EstimationTransformerVideo

🎯 What it does: Proposed a two-step self-supervised transformer-based model for simultaneously estimating monocular depth and visual odometry.

Semantic-focused Patch Tokenizer with Multi-branch Mixer for Visual Place Recognition

Zhenyu Xu, Jun Cheng

RetrievalConvolutional Neural NetworkTransformerImage

🎯 What it does: Propose a vision token-guided VPR framework that includes a semantic-focused patch tokenizer and a multi-branch mixer.

Semantically Guided Feature Matching for Visual SLAM

Oguzhan Ilter, Dániel Baráth

Pose EstimationAutonomous DrivingSimultaneous Localization and Mapping

🎯 What it does: Proposes an algorithm that leverages semantic information to enhance feature matching and loop closure detection in visual SLAM.

Semi-autonomous surface-tracking tasks using omnidirectional mobile manipulators

Carlos Suarez Zapico, M. S. Erden

Robotic Intelligence

🎯 What it does: Proposes a force tracking control scheme for omnidirectional mobile manipulators, enabling human operators to teleoperate the mobile base and end-effector to complete physical interaction tasks with unknown surface geometries.

Semi-Supervised Learning for Visual Bird’s Eye View Semantic Segmentation

Junyu Zhu, Yong Liu

SegmentationImage

🎯 What it does: A visual bird's-eye-view (BEV) semantic segmentation framework based on semi-supervised learning is proposed, which enhances model performance by leveraging unlabeled images through consistency loss and feature constraints, and introduces a conjoint rotation data augmentation method to maintain geometric relationships between front-view and BEV.

Sense in Motion with Belief Clustering: Efficient Gas Source Localization with Mobile Robots

Wanting Jin, A. Martinoli

Robotic IntelligenceTime Series

🎯 What it does: Studied continuous sampling methods for mobile robots in gas source localization tasks, evaluated the impact of gas dynamics and sensor characteristics on localization performance, and proposed a path planning method based on full information.

Sensor-based Multi-Robot Coverage Control with Spatial Separation in Unstructured Environments

Xinyi Wang, Ben M. Chen

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: A decentralized Voronoi coverage control method is proposed, which utilizes active sensing and local sensing information in multi-robot systems to generate collision-free Voronoi regions in irregular environments. It combines deadlock-aware guidance maps with gradient-optimized centroid Voronoi coverage strategies to improve task execution efficiency.

Sensorized Soft Skin for Dexterous Robotic Hands

Jana Egli, Robert K. Katzschmann

Robotic Intelligence

🎯 What it does: Equipping a tendon-driven humanoid hand skeleton with soft, sensor-equipped tactile skin using multi-material 3D printing

Sensorless Estimation of Contact Using Deep-Learning for Human-Robot Interaction

Shilin Shan, Quang Pham

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Propose a sensor-agnostic contact estimation method based on deep learning, utilizing a long-term memory scheme to achieve dynamic identification and precisely approximate static hysteresis;

SeqTrack3D: Exploring Sequence Information for Robust 3D Point Cloud Tracking

Yu Lin, Zheng Fang

Object TrackingAutonomous DrivingPoint Cloud

🎯 What it does: Propose SeqTrack3D, which employs a sequence-to-sequence tracking paradigm, combining historical point clouds and bounding box sequences to capture target motion across consecutive frames.

SEQUEL: Semi-Supervised Preference-based RL with Query Synthesis via Latent Interpolation

Daniel Marta, Iolanda Leite

Reinforcement Learning from Human FeedbackReinforcement LearningAuto EncoderSequential

🎯 What it does: Propose a method for synthesizing queries in semi-supervised preference reinforcement learning through a latent variational autoencoder (VAE), enhancing sample efficiency and generating queries highly consistent with human labels without requiring additional human intervention, further improving the generalization ability of the reward function.

Sequential Manipulation of Deformable Linear Object Networks with Endpoint Pose Measurements using Adaptive Model Predictive Control

T. Toner, Kira Barton

Robotic IntelligenceSupervised Fine-TuningSequential

🎯 What it does: This study addresses the single-arm manipulation of rigidly deformable linear object networks (DLON), achieving stateless estimation control by outputting endpoint pose. A neural network model was trained to verify the existence of input-output dynamics, employing polynomial approximation combined with rigid body dynamics models and online data-driven residuals to construct a composite model. Based on this model, an adaptive model predictive controller was developed to complete DLON installation tasks (including simulation and real vehicle harness installation).

Sequential Trajectory Optimization for Externally-Actuated Modular Manipulators with Joint Locking

Jaeu Choe, Dongjun Lee

OptimizationRobotic Intelligence

🎯 What it does: Proposes a novel trajectory planning method for externally driven modular manipulators (EAMM), utilizing joint locking/unlocking functions to balance payload capacity and flexibility.

SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learning

Jianlan Luo, Sergey Levine

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed a sample-efficient robot reinforcement learning software suite called SERL, which includes deep offline RL algorithms, reward and environment reset methods, high-quality controllers, and example tasks

SG-Bot: Object Rearrangement via Coarse-to-Fine Robotic Imagination on Scene Graphs

Guangyao Zhai, Benjamin Busam

GenerationRobotic IntelligenceGraph Neural NetworkWorld ModelGraph

🎯 What it does: Proposes the SG-Bot framework, which realizes object rearrangement through a coarse-to-fine scene graph imagination approach.

SG-RoadSeg: End-to-End Collision-Free Space Detection Sharing Encoder Representations Jointly Learned via Unsupervised Deep Stereo

Zhiyuan Wu, Rui Fan

Depth EstimationAutonomous DrivingConvolutional Neural NetworkImage

🎯 What it does: Proposed an end-to-end collision-free space detection network called SG-RoadSeg, which utilizes RGB images and unsupervised stereo matching to jointly learn shared encoder representations, achieving collision space segmentation without relying on additional 3D sensors.

SGCalib: A Two-stage Camera-LiDAR Calibration Method Using Semantic Information and Geometric Features

Zhipeng Lin, Yuhan Zhu

Object DetectionPose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingImageMultimodalityPoint Cloud

🎯 What it does: Propose an online two-stage camera-LiDAR extrinsic calibration method that utilizes semantic information and geometric features for calibration.

Shadow-Based 3D Pose Estimation of Intraocular Instrument Using Only 2D Images

Junjie Yang, I. M. A. N. Fellow

Pose EstimationImageBiomedical Data

🎯 What it does: Proposes a shadow-based 3D pose estimation method using only 2D microscope images, which real-time acquires the tip position of intraocular instruments using a spherical eye model with a 12mm radius.

Shaping Social Robot to Play Games with Human Demonstrations and Evaluative Feedback

Chuanxiong Zheng, Guangliang Li

Robotic IntelligenceReinforcement Learning from Human Feedback

🎯 What it does: Enable the social robot Haru to mimic gaming strategies in real-time two-player games through human player demonstration trajectories and evaluation feedback.

Shared Autonomy via Variable Impedance Control and Virtual Potential Fields for Encoding Human Demonstrations*

Shail V Jadav, Dongheui Lee

Safty and PrivacyRobotic Intelligence

🎯 What it does: Proposed and implemented a framework for complex human-robot collaboration tasks (e.g., furniture co-manufacturing), focusing on task encoding and safe compliant execution.

Short term after-effects of small force fields applied by an upper-limb exoskeleton on inter-joint coordination

Océane Dubois, N. Jarrassé

Robotic IntelligenceBiomedical Data

🎯 What it does: Investigated the short-term aftereffects of low-amplitude force fields applied by upper limb exoskeletons on inter-joint coordination in non-symptomatic subjects

SiLVR: Scalable Lidar-Visual Reconstruction with Neural Radiance Fields for Robotic Inspection

Yifu Tao, Maurice F. Fallon

Robotic IntelligenceNeural Radiance FieldSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: By fusing LiDAR and visual data, large-scale reconstruction is achieved based on neural radiance fields (NeRF), generating geometrically accurate and photo-realistic textured images.

Sim-to-Real Grasp Detection with Global-to-Local RGB-D Adaptation

Haoxiang Ma, Di Huang

Object DetectionDomain AdaptationContrastive LearningImageBenchmark

🎯 What it does: Proposes a simulation-to-real domain adaptation method for RGB-D grasp detection, achieving domain adaptation through self-supervised rotation pre-training and a global-local alignment pipeline.

Sim-to-Real Learning for Humanoid Box Loco-Manipulation

Jeremy Dao, Alan Fern

Domain AdaptationRobotic IntelligenceReinforcement Learning

🎯 What it does: Trained and deployed a sim-to-real reinforcement learning controller, enabling the bipedal robot Digit to carry boxes of varying sizes, weights, and initial poses while maintaining balance.

Sim-to-real Object Pose Estimation for Random Bin Picking

Boyoung Kim, Junhong Min

Data SynthesisPose EstimationDomain AdaptationPoint Cloud

🎯 What it does: Proposes a random bin picking object pose estimation method that is trained using only simulated data and can be directly applied to real environments without additional adaptation; simultaneously constructs a fully automated synthetic dataset based on a physics simulator and 3D CAD models, and generates annotations for instance segmentation and pose estimation.

Sim-to-Real Robotic Sketching using Behavior Cloning and Reinforcement Learning

Biao Jia, Dinesh Manocha

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Train a painting strategy effective in both simulated and real environments using behavior cloning and reinforcement learning, constructing a robot drawing system based on the UltraArm robotic arm and RealSense D415 camera;

Sim2Real Bilevel Adaptation for Object Surface Classification using Vision-Based Tactile Sensors

Ga Caddeo, Lorenzo Natale

ClassificationData SynthesisPose EstimationDomain AdaptationDiffusion modelImage

🎯 What it does: Align a small number of real-world images with simulated images using diffusion models, train a surface classifier for tactile sensors via adversarial feature alignment, and evaluate performance on real-world tactile images.

Sim2Real Manipulation on Unknown Objects with Tactile-based Reinforcement Learning

Entong Su, Xiaolong Wang

Robotic IntelligenceReinforcement LearningMultimodality

🎯 What it does: Using visual-tactile perception input in a physics simulation environment for reinforcement learning to train grasp and manipulation strategies for diverse objects, transferring them to real robots; simultaneously studying different tactile representations to alleviate Sim2Real issues.

SIMMF: Semantics-aware Interactive Multiagent Motion Forecasting for Autonomous Vehicle Driving

Vidyaa Krishnan Nivash, A. H. Qureshi

Autonomous Driving

🎯 What it does: Propose a semantic-aware interactive multi-agent motion prediction method (SIMMF) for multi-agent trajectory prediction in autonomous driving.

Simulation and Experimental Validation of an Autonomous Perching and Takeoff Method for a Multirotor UAV on Vertical Surfaces using a Suction Cup

Bruno Chapdelaine, B. Monsarrat

OptimizationRobotic Intelligence

🎯 What it does: Conducted simulation and experimental verification of autonomous landing and takeoff methods for suction cup-based multi-rotor drones on vertical surfaces.

Simulation Modeling of Highly Dynamic Omnidirectional Mobile Robots Based on Real-World Data

Marvin Wiedemann, Sören Kerner

Robotic IntelligenceWorld ModelSequential

🎯 What it does: Simulate a high-dynamic omnidirectional logistics robot and use motion capture data to achieve matching and optimization between simulation and the physical machine.

Simultaneous Estimation of Shape and Force along Highly Deformable Surgical Manipulators Using Sparse FBG Measurement

Yiang Lu, Yun-hui Liu

Robotic IntelligenceTime SeriesBiomedical Data

🎯 What it does: Propose a sparse strain measurement based on single-core FBG fiber, using a data-driven approach to simultaneously estimate shape and external force.

Simultaneous Time Synchronization and Mutual Localization for Multi-robot System

Xiangyong Wen, Fei Gao

OptimizationRobotic Intelligence

🎯 What it does: Integrate time synchronization into mutual localization to jointly recover time offsets and relative poses between robots;

Singularity Analysis of Kinova’s Link 6 Robot Arm via Grassmann Line Geometry

Milad Asgari, Clément Gosselin

Robotic Intelligence

🎯 What it does: Studied the singularities of the Kinova Link 6 robotic arm, using Grassmann line geometry to analyze all screw dependency cases, deriving 12 singular configurations and providing seven simple geometric conditions.

Singularity-Robust Prioritized Whole-Body Tracking and Interaction Control With Smooth Task Transitions

Xuwei Wu, Alexander Dietrich

Robotic Intelligence

🎯 What it does: Propose a singularity-robust whole-body control framework that achieves smooth task switching while maintaining strict task priority.

SKD-Net: Spectral-based Knowledge Distillation in Low-Light Thermal Imagery for robotic perception

Aniruddh Sikdar, Suresh Sundaram

SegmentationKnowledge DistillationRobotic IntelligenceImageMultimodality

🎯 What it does: Proposes a spectral-based knowledge distillation architecture named SKD-Net to enhance the performance of semantic segmentation models under low-light thermal imaging scenarios when a modality is missing (only using IR images).

Skill Learning in Robot-Assisted Micro-Manipulation Through Human Demonstrations with Attention Guidance

Yujian An, Guang-Zhong Yang

OptimizationRobotic Intelligence

🎯 What it does: Learn micro-operation skills through human demonstrations and eye-tracking guided attention, train neural networks for visual ROI segmentation, derive optimized action strategies, and apply them to robot automation for micro-assembly tasks.

Skill Transfer for Temporal Task Specification

J. Liu, Stefanie Tellex

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes an LTL-based zero-shot transfer algorithm called LTL-Transfer, which can safely satisfy various novel LTL task specifications using task-agnostic skills learned during training without violating safety constraints.

SKT-Hang: Hanging Everyday Objects via Object-Agnostic Semantic Keypoint Trajectory Generation

Chia-Liang Kuo, Yi-Ting Chen

Robotic Intelligence

🎯 What it does: Studied hanging multiple grasped objects on diverse supports and proposed a general semantic keypoint trajectory (SKT) representation and shape-conditioned trajectory deformation network (SCTDN) to generate trajectories.

SLAM Based on Camera-2D LiDAR Fusion

Guoyu Lu

Robotic IntelligenceSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposes a SLAM system based on the fusion of a camera and a single 2D LiDAR, and dynamically adjusts scale uncertainty through a deep learning framework and self-supervised network.

SLCF-Net: Sequential LiDAR-Camera Fusion for Semantic Scene Completion using a 3D Recurrent U-Net

Helin Cao, Sven Behnke

SegmentationDepth EstimationAutonomous DrivingConvolutional Neural NetworkRecurrent Neural NetworkImagePoint CloudSequential

🎯 What it does: Propose a serial fusion SLCF-Net that integrates LiDAR and camera data for semantic scene completion tasks, jointly estimating missing geometric and semantic information of the scene.

SliceIt! - A Dual Simulator Framework for Learning Robot Food Slicing

C. C. Beltran-Hernandez, Masashi Hamaya

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed the SliceIt! framework, integrating high-fidelity cutting simulation with robot simulation, training compliant control policies through reinforcement learning to achieve a full workflow from limited real cutting data to real robot deployment.

SLIM: Skill Learning with Multiple Critics

David Emukpere, J. Renders

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a multi-evaluator learning framework named SLIM for self-supervised skill discovery in robotic manipulation

SlotGNN: Unsupervised Discovery of Multi-Object Representations and Visual Dynamics

Alireza Rezazadeh, Changhyun Choi

Representation LearningRobotic IntelligenceGraph Neural NetworkTransformerImage

🎯 What it does: Propose a framework based on SlotTransport and SlotGNN that can learn multi-object representations and dynamics prediction from RGB images and robot interactions under unsupervised conditions.

SM3: Self-supervised Multi-task Modeling with Multi-view 2D Images for Articulated Objects

Haowen Wang, Jian Tang

Pose EstimationImageMultimodality

🎯 What it does: Proposes the SM3 self-supervised interactive perception method, which models and identifies movable parts of joints before and after interaction using multi-view RGB images, and infers their rotational joint parameters.

SmartCooper: Vehicular Collaborative Perception with Adaptive Fusion and Judger Mechanism

Yuang Zhang, Yuguang Fang

Autonomous DrivingImage

🎯 What it does: Proposes the SmartCooper framework, combining communication resource optimization and discriminative mechanisms to achieve adaptive fusion for vehicle cooperative perception;

Smooth Computation without Input Delay: Robust Tube-Based Model Predictive Control for Robot Manipulator Planning

Qie Sima, Jianwei Zhang

OptimizationRobotic Intelligence

🎯 What it does: Propose a robust pipeline MPC method for robotic manipulation that predicts subsequent states in advance to reduce OCP computation delay and improve response speed.

Snake Robot with Tactile Perception Navigates on Large-scale Challenging Terrain

Shuo Jiang, Lawson L. S. Wong

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a framework that integrates tactile sensing into the locomotion control of snake-like robots to enhance their adaptability to various terrains.

SocialGAIL: Faithful Crowd Simulation for Social Robot Navigation

Bo Ling, Weiwei Wu

Autonomous DrivingReinforcement Learning from Human FeedbackGraph Neural NetworkReinforcement LearningGenerative Adversarial Network

🎯 What it does: Propose a data-driven crowd simulation method named SocialGAIL based on generative adversarial imitation learning, aiming to more realistically reproduce pedestrian navigation behavior in crowded environments;

Soft Acoustic End-effector

Zhiyuan Zhang, Daniel Ahmed

Robotic IntelligenceUltrasound

🎯 What it does: Proposed and demonstrated a soft acoustic end-effector that can modulate the sound field and manipulate microparticles by adjusting the position and orientation of the PZT.

Soft Hand Extension Glove with Thumb Abduction and Extension Assistance

Disheng Xie, Raymond Kai‐yu Tong

Robotic Intelligence

🎯 What it does: Developed a fully wearable soft hand extension glove based on the X-pouch and strap system, capable of achieving finger extension, thumb abduction, and extension to meet the needs of patients with high MAS scores.

SOL: A Compact, Portable, Telescopic, Soft-Robotic Sun-Tracking Mechanism for Improved Solar Power Production

Bryan Busby, Minas Liarokapis

Robotic Intelligence

🎯 What it does: Proposed and experimentally verified a compact, portable soft robotic dual-axis solar tracking mechanism SOL to enhance photovoltaic panel efficiency.

Solving Sequential Manipulation Puzzles by Finding Easier Subproblems

Svetlana Levit, Marc Toussaint

OptimizationRobotic Intelligence

🎯 What it does: Propose solving the multi-object sequential manipulation challenge by searching for a sequence of simpler grasp-place subproblems, and combine heuristic forward search with optimization-based task and motion planning solvers.

SONIC: Sonar Image Correspondence using Pose Supervised Learning for Imaging Sonars

Samiran Gode, Michael Kaess

Pose EstimationImage

🎯 What it does: Proposed the SONIC (SONar Image Correspondence) network, which achieves feature correspondence for underwater sonar images through pose-supervised learning, maintaining robustness under viewpoint changes;

SpaceHopper: A Small-Scale Legged Robot for Exploring Low-Gravity Celestial Bodies

Alexander Spiridonov, Marco Hutter

Robotic IntelligenceReinforcement LearningPhysics Related

🎯 What it does: A three-legged small robot named SpaceHopper was developed for mobility exploration on low-gravity celestial bodies, achieving body reorientation during flight through leg movement.

SPADES: A Realistic Spacecraft Pose Estimation Dataset using Event Sensing

A. Rathinam, D. Aouada

Data SynthesisPose EstimationDomain AdaptationBenchmarkPhysics Related

🎯 What it does: Proposed the SPADES dataset, which includes real event data obtained in the laboratory and simulated event data with the same camera intrinsic parameters, and introduced a new image-based event representation method and effective data filtering strategies, conducting multi-dimensional baseline evaluations.

Spatial Assisted Human-Drone Collaborative Navigation and Interaction through Immersive Mixed Reality

Luca Morando, Giuseppe Loianno

Robotic IntelligenceSimultaneous Localization and MappingMultimodality

🎯 What it does: A novel remote immersive framework is proposed, leveraging mixed reality to promote collaboration between humans and drones at both cognitive and physical levels, and verifying collaborative planning and exploration tasks through bidirectional spatial perception and multimodal virtual-physical interaction.

Spatio-Temporal Correspondence Estimation of Growing Plants by Hausdorff Distance based Skeletonization for Organ Tracking

Sharmistha B. Pandey, Ayan Chaudhury

Object TrackingPoint CloudAgriculture Related

🎯 What it does: Proposes a skeletonization method based on Hausdorff distance and an improved BFS algorithm, and estimates correspondence between skeleton graphs using cosine similarity to achieve tracking of plant organs in spatiotemporal point cloud sequences.

SpawnNet: Learning Generalizable Visuomotor Skills from Pre-trained Network

Xingyu Lin, P. Abbeel

Representation LearningRobotic IntelligenceConvolutional Neural NetworkReinforcement LearningImage

🎯 What it does: The study investigates the generalization ability of pre-trained visual representations at the class level, and proposes SpawnNet's two-stream architecture, which fuses pre-trained multi-layer representations into independent networks to learn robust strategies.

SPCGC: Scalable Point Cloud Geometry Compression for Machine Vision

Liang Xie, Ge Li

ClassificationSegmentationCompressionPoint Cloud

🎯 What it does: Propose an expandable point cloud geometry compression framework, incorporating base layer geometric data and enhanced layer semantic-guided residual data, while introducing classification and segmentation losses from downstream tasks into Rate-Distortion optimization;

Specifying and Monitoring Safe Driving Properties with Scene Graphs

Felipe Toledo, Matthew B. Dwyer

Autonomous DrivingSafty and PrivacyGraph Neural NetworkGraph

🎯 What it does: Built and implemented a framework that can extract scene graphs (SG) from sensor inputs and construct propositions using domain-specific languages and temporal logic to monitor the safety driving properties of autonomous vehicles.

Spined Torso Renders Advanced Mobility for Quadrupedal Locomotion

Jichao Wang, Shiwu Zhang

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed a quadruped robot named STRAY with a four-degree-of-freedom spine design, achieving dynamic gaits through the spine using trajectory-based reinforcement learning.

Spline-Interpolated Model Predictive Path Integral Control with Stein Variational Inference for Reactive Navigation

Takato Miura, Susumu Hara

Autonomous DrivingOptimization

🎯 What it does: A reactive navigation method based on Model Predictive Path Integral (MPPI) is proposed, incorporating spline interpolation in its control input sequence and using Stein Variational Gradient Descent (SVGD) to handle multi-solution scenarios.

SPOT: Point Cloud Based Stereo Visual Place Recognition for Similar and Opposing Viewpoints

Spencer Carmichael, K. Skinner

RetrievalSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a method called SPOT for opposite perspective visual place recognition based on stereo visual odometry structure estimation.

SPOTS: Stable Placement of Objects with Reasoning in Semi-Autonomous Teleoperation Systems

Joonhyung Lee, Sungjoon Choi

Robotic IntelligenceTransformerLarge Language Model

🎯 What it does: Propose a semi-autonomous remote control system that integrates real-time simulation-based physical stability verification with large language model semantic reasoning for stably and reasonably placing objects.

SPRINT: Scalable Policy Pre-Training via Language Instruction Relabeling

Jesse Zhang, Joseph J. Lim

Robotic IntelligenceLarge Language ModelReinforcement Learning

🎯 What it does: Proposes SPRINT, a scalable offline policy pre-training method that significantly reduces the human annotation workload required for pre-training diverse skills.

Square-Root Inverse Filter-based GNSS-Visual-Inertial Navigation

Jun Hu, Guoquan Huang

Autonomous DrivingOptimizationSimultaneous Localization and MappingMultimodality

🎯 What it does: Developed a new GNSS-visual-inertial navigation system (GVINS), named SRI-GVINS, which tightly fuses visual, inertial, and raw GNSS measurements through the Square Root Inverse Sliding Window Filter (SRI-SWF) framework.

Squirrel-inspired Tendon-driven Passive Gripper for Agile Landing

Stanley J. Wang, Hannah S. Stuart

Robotic Intelligence

🎯 What it does: Studied the parameterized design of a passive gripper inspired by koalas for high-impact landings, with a fixed geometric structure and experimental adjustments through varying joint stiffness and contact conditions.

SRFNet: Monocular Depth Estimation with Fine-grained Structure via Spatial Reliability-oriented Fusion of Frames and Events

Tianbo Pan, Lin Wang

Depth EstimationImageMultimodality

🎯 What it does: Propose SRFNet for monocular depth estimation, fusing frame images and event stream data

Stability Analysis of Distance-Angle Leader-Follower Formation Control*

Manao Machida, Masumi Ichien

OptimizationRobotic Intelligence

🎯 What it does: Studied the stability of distance-angle leader-follower formation control, provided necessary and sufficient conditions, and proposed a globally asymptotically stable controller.