arXivSub Start free trial

ICRA 2023 Papers — Page 12

IEEE International Conference on Robotics and Automation · 1341 papers

Self-adaptive Teaching-learning-based Optimizer with Improved RBF and Sparse Autoencoder for Complex Optimization Problems

J. Bi, Mengchu Zhou

OptimizationAuto EncoderBenchmark

🎯 What it does: Proposed an adaptive teaching-learning optimizer (STORA), combining an improved radial basis function (RBF) model and sparse autoencoder to address high-dimensional complex optimization problems.

Self-Entanglement-Free Tethered Path Planning for Non-Particle Differential-Driven Robot

Tong Yang, R. Xiong

Robotic Intelligence

🎯 What it does: Proposed a new mechanism for generating self-entanglement-free paths for traction differential drive robots without omnidirectional rope retraction mechanisms, along with a search-based near-optimal path planning algorithm.

Self-Improving Safety Performance of Reinforcement Learning Based Driving with Black-Box Verification Algorithms

Resul Dagdanov, N. K. Ure

Autonomous DrivingSafty and PrivacyReinforcement Learning

🎯 What it does: Propose a self-improving AI system that enhances the safety performance of reinforcement learning-based autonomous driving through black-box verification methods

Self-supervised Cloth Reconstruction via Action-conditioned Cloth Tracking

Zixuan Huang, David Held

GenerationMesh

🎯 What it does: Propose a self-supervised method to fine-tune a mesh reconstruction model in the real world by generating pseudo labels using action-conditioned cloth tracking.

Self-Supervised Learning of Action Affordances as Interaction Modes

Liquang Wang, Animesh Garg

GenerationRepresentation LearningRobotic IntelligenceAuto EncoderImage

🎯 What it does: Propose a framework for learning affordance interaction patterns using depth sensors in a simulator, defining successful interactions as significant changes in the visual environment, and learning models that can generate these interactions, which can be conditioned on target states.

Self-Supervised Learning of Object Segmentation from Unlabeled RGB-D Videos

Shiyang Lu, Kostas E. Bekris

SegmentationContrastive LearningVideoPoint Cloud

🎯 What it does: Proposed a self-supervised learning system for segmenting rigid objects in RGB images, using unlabeled RGB-D video as training data.

Self-Supervised Monocular Depth Underwater

Shlomi Amitai, T. Treibitz

Depth EstimationVideo

🎯 What it does: Train a self-supervised monocular depth network using subsequent frames and reprojection loss, adding multiple improvements to adapt to underwater environments, achieving the best results.

Self-supervised Multi-frame Monocular Depth Estimation with Pseudo-LiDAR Pose Enhancement

Wenhua Wu, Zhe Liu

Depth EstimationImagePoint Cloud

🎯 What it does: Proposes a monocular multi-frame unsupervised depth estimation framework called PLPE-Depth, which estimates the pose of adjacent frames using pseudo LiDAR reconstructed from depth maps and re-estimates depth between image poses and pseudo LiDAR poses to enhance the coupling between depth and pose;

Semantic Keypoint Extraction for Scanned Animals using Multi-Depth-Camera Systems

R. Falque, A. Alempijevic

Pose EstimationPoint Cloud

🎯 What it does: Propose a framework that transforms semantic keypoint extraction into a regression problem based on the distance between keypoints and the rest of the point cloud, achieved through radial basis function (RBF) mapping on the point cloud manifold combined with an encoder-decoder network; simultaneously, design a data augmentation scheme for multi-depth camera systems addressing extrinsic noise and frame loss, and explore efficient non-rigid deformation methods.

Semantic Mapping with Confidence Scores through Metric Embeddings and Gaussian Process Classification

Jungseok Hong, Volkan Isler

Simultaneous Localization and MappingImage

🎯 What it does: Proposes a mapping method that integrates semantic information with shape completion inferred from RGBD images, and calculates the confidence score of its predictions.

Semantic-SuPer: A Semantic-aware Surgical Perception Framework for Endoscopic Tissue Identification, Reconstruction, and Tracking

Shan Lin, Michael C. Yip

ClassificationObject TrackingSegmentationConvolutional Neural NetworkVideoBiomedical Data

🎯 What it does: Introduces a comprehensive surgical perception framework called Semantic-SuPer for the identification, 3D reconstruction, and tracking of tissues in endoscopic videos.

Semantics-aware Exploration and Inspection Path Planning

M. Dharmadhikari, K. Alexis

Robotic Intelligence

🎯 What it does: Proposed a semantics-aware autonomous exploration and detection path planning strategy

Semi-autonomous robotic control of a self-shaping cochlear implant

D. Bautista-Salinas, F. Baena

Robotic Intelligence

🎯 What it does: A vision-based collaborative method was used to control the semi-autonomous robotic insertion of an adaptive cochlear implant, successfully inserting a pre-curved thermoplastic electrode array.

Sensor Localization by Few Distance Measurements via the Intersection of Implicit Manifolds

Michael M. Bilevich, D. Halperin

Robotic Intelligence

🎯 What it does: A method is proposed to locate the unknown position and orientation of a sensor by utilizing 2 to 6 distance measurements and the robot's execution of predetermined local motions in a known environment; the method achieves localization by intersecting implicitly defined 2D manifolds in a 3D configuration space, and can reduce the set of possible poses to curves or even point sets in the plane workspace.

Seq2Seq Imitation Learning for Tactile Feedback-based Manipulation

Wenyan Yang, Joni-Kristian Kämäräinen

Robotic IntelligenceSequential

🎯 What it does: Propose a Seq2Seq-based imitation learning framework for solving partially observable, contact-rich manipulation tasks under tactile feedback.

Sequence-Agnostic Multi-Object Navigation

Nandiraju Gireesh, M. Krishna

Autonomous DrivingReinforcement LearningImageBenchmark

🎯 What it does: Proposes a deep reinforcement learning framework based on the actor-critic architecture for multi-object navigation tasks involving unordered sequences.

Sequential Bayesian Optimization for Adaptive Informative Path Planning with Multimodal Sensing

Joshua Ott, Mykel J. Kochenderfer

OptimizationMultimodality

🎯 What it does: Proposes the problem of adaptive information path planning under multi-modal perception (AIPPMS), formulating it as a belief Markov decision process with Gaussian process beliefs; solving the problem using sequential Bayesian optimization and online planning methods.

Sequential Stochastic Multi-Task Assignment for Multi-Robot Deployment Planning

C. Mitchell, Geoffrey A. Hollinger

OptimizationRobotic Intelligence

🎯 What it does: Proposed the MARP algorithm, a multi-robot multi-task sequence random allocation algorithm based on simulation optimization

SGDViT: Saliency-Guided Dynamic Vision Transformer for UAV Tracking

L. Yao, Junjie Ye

Object TrackingTransformerVideo

🎯 What it does: Proposed a Significance-Guided Dynamic Vision Transformer (SGDViT) for UAV tracking

SGPT: The Secondary Path Guides the Primary Path in Transformers for HOI Detection

Sixian Chan, Cong Bai

Object DetectionTransformerImage

🎯 What it does: Proposed the SGPT method, which employs a secondary path to guide the main path for HOI detection.

SGTM 2.0: Autonomously Untangling Long Cables using Interactive Perception

K. Shivakumar, Ken Goldberg

Robotic Intelligence

🎯 What it does: Utilizes interactive perception technology to automatically dismantle a 3-meter-long cable, completing the rope disentanglement through sliding and grasping actions

SHAIL: Safety-Aware Hierarchical Adversarial Imitation Learning for Autonomous Driving in Urban Environments

Arec L. Jamgochian, M. Kochenderfer

Autonomous Driving

🎯 What it does: Learn a high-level policy using safety-aware hierarchical adversarial imitation learning (SHAIL), selecting from a set of low-level controllers to achieve safe and human-like driving decisions in urban roundabout scenarios.

Shape visual servoing of a tether cable from parabolic features

Lev Smolentsev, F. Chaumette

Robotic IntelligencePoint Cloud

🎯 What it does: Proposes a method for visual servoing of suspended rope deformation using RGB-D camera visual features, and provides the interaction matrix, real-time feature extraction algorithm, and experimental validation.

Shared Control of Assistive Robots through User-intent Prediction and Hyperdimensional Recall of Reactive Behavior

Alisha Menon, J. Rabaey

Robotic IntelligenceMultimodality

🎯 What it does: Proposes a user-adaptive multi-layer shared control scheme that utilizes brain-inspired high-dimensional computing (HDC) for classifying and recalling reactive behaviors in assistive robots, achieving execution control based on user goals, and implementing shared control through multimodal activity data for identifying recent behaviors, predicting the next action, and integrating HDC recall.

SHINE-Mapping: Large-Scale 3D Mapping Using Sparse Hierarchical Implicit Neural Representations

Xingguang Zhong, C. Stachniss

OptimizationComputational EfficiencyRepresentation LearningNeural Radiance FieldSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Utilizes sparse hierarchical implicit neural representations to construct large-scale 3D reconstructions based on 3D LiDAR measurements, and designs an incremental mapping system to address the forgetting problem in continual learning.

Show me What you want: Inverse Reinforcement Learning to Automatically Design Robot Swarms by Demonstration

Ilyes Gharbi, M. Birattari

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Developed the Demo-Cho method, which utilizes inverse reinforcement learning and automated modular design to automatically generate swarm control software for robots from demonstrations.

Shunted Collision Avoidance for Multi-UAV Motion Planning with Posture Constraints

Gang Xu, Jian Yang

Autonomous Driving

🎯 What it does: Study motion planning for fixed-wing UAVs under attitude constraints and address the multi-solution symmetric scenario

SIERRA: A Modular Framework for Accelerating Research and Improving Reproducibility

John Harwell, Maria L. Gini

Computational Efficiency

🎯 What it does: Proposed and implemented the SIERRA framework, which automatically generates experiments, executes experiments, and processes results to generate charts and videos, reducing the manual configuration burden on researchers.

Sim-and-Real Reinforcement Learning for Manipulation: A Consensus-based Approach

Wenxing Liu, J. Carrasco

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a consensus-based Sim-And-Real deep reinforcement learning algorithm (CSAR) for robotic arm grasping and placing tasks, with training conducted simultaneously in simulation environments and the real world.

Sim-to-Real Policy and Reward Transfer with Adaptive Forward Dynamics Model

Rongshun Juan, Guangliang Li

Domain AdaptationReinforcement LearningWorld Model

🎯 What it does: Proposes a Progressive Policy Transfer and Adaptive Dynamics Model (PPTADM) to achieve simulation-to-reality transfer learning, and improves learning efficiency and performance in real environments through progressive neural networks and forward dynamics models.

Sim-to-Real Transfer for Quadrupedal Locomotion via Terrain Transformer

Hang Lai, Jun Wang

Domain AdaptationRobotic IntelligenceTransformer

🎯 What it does: Proposed a high-capacity Transformer model called Terrain Transformer (TERT) and designed a two-stage training framework (offline pre-training and online correction) for simulation-to-real transfer of quadruped robots across different terrains.

Sim2Real2: Actively Building Explicit Physics Model for Precise Articulated Object Manipulation

Liqian Ma, Rui Chen

Robotic IntelligenceWorld ModelPoint CloudPhysics Related

🎯 What it does: Propose the Sim2Real2 framework to achieve precise manipulation of unseen articulated objects in real environments without human demonstrations.

Simple-BEV: What Really Matters for Multi-Sensor BEV Perception?

Adam W. Harley, Katerina Fragkiadaki

Autonomous DrivingMultimodality

🎯 What it does: Analyze the design and training protocol of multi-sensor BEV perception models, clarify the impact of batch size and input resolution on performance, and demonstrate that radar data can significantly improve performance, narrowing the gap between cameras and LiDAR systems.

Simplified Motor Primitives for Gait Symmetrization: Pilot Study with an Active Hip Orthosis

Henri Laloyaux, R. Ronsse

Robotic IntelligenceBiomedical Data

🎯 What it does: Designed and verified an algorithm utilizing motor primitives to achieve gait symmetrization, with preliminary validation conducted on a single simulated hemiplegic gait subject using a hip exoskeleton.

Simplifying Aerial Manipulation Using Intentional Collisions

Mark Nail, R. Gillespie

Robotic Intelligence

🎯 What it does: Developed a UAV aerial manipulation method utilizing intentional collisions, designed a Velocity Matching controller, proposed a flight envelope assessment method before collision, and verified its recoverability through simulation and experiments.

Simultaneous Tactile Estimation and Control of Extrinsic Contact

Sangwoon Kim, Alberto Rodriguez

Robotic IntelligenceMultimodality

🎯 What it does: Propose a factor graph framework that utilizes tactile feedback to simultaneously estimate and control external contact, enabling manipulation tasks requiring delicate force application and precise contact maneuvers, such as balancing unknown objects on a thin rod.

Skill-based Robot Programming in Mixed Reality with Ad-hoc Validation Using a Force-enabled Digital Twin

J. Krieglstein, Werner Kraus

Robotic Intelligence

🎯 What it does: Developed a mixed reality system that intuitively displays the geometric constraints of high-level skills in real robot working environments, and enables instant verification of programs through holographic digital twins and force control simulation; the system also provides a user interface for creation and debugging, and was evaluated in a top-hat track installation task.

SLAMER: Simultaneous Localization and Map-Assisted Environment Recognition

Naoki Akai

Robotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a SLAMER method that integrates map prediction with sensor recognition to achieve simultaneous localization and environment recognition.

SLAMesh: Real-time LiDAR Simultaneous Localization and Meshing

Jianyuan Ruan, Yuxiang Sun

Autonomous DrivingComputational EfficiencySimultaneous Localization and MappingPoint CloudMesh

🎯 What it does: Proposed a CPU-only real-time LiDAR SLAM system that can simultaneously construct a grid map and perform localization.

Slice Transformer and Self-supervised Learning for 6DoF Localization in 3D Point Cloud Maps

Muhammad Ibrahim, Ajmal Saeed Mian

Pose EstimationRepresentation LearningTransformerPoint Cloud

🎯 What it does: Propose a self-supervised learning method that reorganizes slices of outdoor 360° LiDAR scans using Transformer to achieve 6DoF localization; simultaneously conduct experiments on the newly constructed large-scale LiDAR map Perth-WA, and apply this method to downstream object classification tasks.

SLURP! Spectroscopy of Liquids Using Robot Pre-Touch Sensing

Nathaniel Hanson, Zackory Erickson

ClassificationRobotic IntelligenceTabularPhysics Related

🎯 What it does: Propose integrating visible to near-infrared (VNIR) reflectance spectroscopy technology into robot end-effectors to identify liquids and granular media within unknown containers.

SM/VIO: Robust Underwater State Estimation Switching Between Model-based and Visual Inertial Odometry

Bharat Joshi, Ioannis M. Rekleitis

Pose EstimationSimultaneous Localization and Mapping

🎯 What it does: To address the robustness issue of visual-inertial state estimation in underwater environments, this paper proposes an SM/VIO estimation framework that switches between model-driven and visual-inertial odometry (VIO).

Small-shot Multi-modal Distillation for Vision-based Autonomous Steering

Yu Shen, Ming-Chyuan Lin

Autonomous DrivingKnowledge DistillationMultimodality

🎯 What it does: Proposed a few-shot multimodal distillation network called AMD-S-Net for visual navigation tasks in autonomous driving.

SmartRainNet: Uncertainty Estimation For Laser Measurement in Rain

Chen Zhang, D. Rus

Autonomous DrivingPoint Cloud

🎯 What it does: Proposed an uncertainty assessment method for each laser point in 3D LiDAR measurements during rainy conditions.

Socially Fair Coverage Control

Matthew Malencia, Vijay R. Kumar

OptimizationTabular

🎯 What it does: Developed a coverage control algorithm focused on social fairness, aiming to minimize the maximum coverage cost between different groups, and constructed a gradient controller to balance fairness and average performance by combining Voronoi iteration with log-exponential sum approximation to tackle non-differentiable objectives.

Soft Sensing Skins for Arbitrary Objects: An Automatic Framework

Sonja Groß, Sami Haddadin

Robotic Intelligence

🎯 What it does: Proposes a partially automated framework for designing and customizing silicone-based skin-like tactile sensors for objects of arbitrary shapes, and evaluates the performance of the stretchable and contact sensors in position control, grasping, and manipulation scenarios.

SoLo T-DIRL: Socially-Aware Dynamic Local Planner based on Trajectory-Ranked Deep Inverse Reinforcement Learning

Yifan Xu, Maani Ghaffari

Autonomous DrivingReinforcement Learning

🎯 What it does: Proposes a social robot navigation framework based on deep inverse reinforcement learning, enabling social navigation in dynamic crowded environments

SonicFinger: Pre-touch and Contact Detection Tactile Sensor for Reactive Pregrasping

Siddharth Rupavatharam, Volkan Isler

Robotic IntelligenceTime Series

🎯 What it does: Proposes a tactile sensor called SonicFinger based on acoustic glory for pre-touch and contact detection, demonstrating its application in robot pre-grasp pose calibration and successful grasping.

Sonicverse: A Multisensory Simulation Platform for Embodied Household Agents that See and Hear

Ruohan Gao, Jiajun Wu

Data SynthesisDomain AdaptationRobotic IntelligenceSimultaneous Localization and MappingWorld ModelVideoMultimodalityAudio

🎯 What it does: Propose Sonicverse, a multisensory simulation platform that can real-time render audio in 3D environments and train home agents capable of both visual and auditory perception.

Source-free Unsupervised Domain Adaptation for 3D Object Detection in Adverse Weather

Deepti Hegde, Vishal M. Patel

Object DetectionData SynthesisDomain AdaptationPoint Cloud

🎯 What it does: Proposed an uncertainty-aware mean teacher framework for source-free unsupervised domain adaptation of 3D object detection networks under adverse weather conditions.

Spatial-Temporal-Aware Safe Multi-Agent Reinforcement Learning of Connected Autonomous Vehicles in Challenging Scenarios

Zhili Zhang, Fei Miao

Autonomous DrivingSafty and PrivacyGraph Neural NetworkTransformerReinforcement Learning

🎯 What it does: Proposes a restricted multi-agent reinforcement learning framework combined with parallel safety shielding for enhancing the safety and efficiency of connected autonomous vehicles (CAV) in challenging driving scenarios involving non-connected hazardous vehicles.

Speeding Up Assembly Sequence Planning Through Learning Removability Probabilities

Alexander Cebulla, T. Kröger

OptimizationComputational EfficiencyGraph Neural NetworkGraph

🎯 What it does: Accelerate assembly sequence planning by predicting disassembly probability through deep learning and optimizing the testing order of parts in assembly-disassembly (AbD).

Spherical Cubic Blends: $\mathcal{C}^{2}$-Continuous, Zero-Clamped, and Time-Optimized Interpolation of Quaternions

J. Wittmann, D. Rixen

OptimizationRobotic Intelligence

🎯 What it does: Proposed a C² continuous, zero-hold interpolation scheme for quaternions, and modified the existing SQUAD and SPB quaternion interpolation methods to also possess C² continuity and zero-hold characteristics, achieving synchronous and efficient motion planning for given key points.

SphNet: A Spherical Network for Semantic Pointcloud Segmentation

Lukas Bernreiter, César Cadena

SegmentationConvolutional Neural NetworkPoint Cloud

🎯 What it does: Proposes a framework for semantic point cloud segmentation using spherical convolutional neural networks, supporting data from different LiDAR sensors.

SRI-Graph: A Novel Scene-Robot Interaction Graph for Robust Scene Understanding

D. Yang, E. Steinbach

Robotic IntelligenceGraph Neural NetworkImage

🎯 What it does: Proposes a method for robust scene understanding using a scene-robot interaction graph (SRI-Graph) that leverages the known position of a mobile manipulator.

Stable Contact Guaranteeing Motion/Force Control for an Aerial Manipulator on an Arbitrarily Tilted Surface

Jeonghyun Byun, H. Kim

Robotic Intelligence

🎯 What it does: A motion/force controller for aerial manipulators was designed to ensure tracking of time-varying motion/force trajectories and maintain system stability when switching between free motion and contact motion.

Stable Station Keeping of Autonomous Sailing Robots via the Switched Systems Approach for Ocean Observation

Weimin Qi, Huihuan Qian

Robotic IntelligencePhysics Related

🎯 What it does: A stable dwell control scheme based on a switched system was developed to enable autonomous sailboats to navigate safely and stably for data collection within a confined observation area.

Stackelberg Games for Learning Emergent Behaviors During Competitive Autocurricula

Boling Yang, Joshua R. Smith

Reinforcement Learning

🎯 What it does: Propose a Stackelberg game-based multi-agent deep deterministic policy gradient algorithm (ST-MADDPG) to address environmental asymmetry and enhance the quality of collaborative evolution strategies.

STAP: Sequencing Task-Agnostic Policies

Christopher Agia, Jeannette Bohg

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposes the Sequencing Task-Agnostic Policies (STAP) framework for training manipulation skills and coordinating their geometric dependencies during planning to address long-horizon tasks;

Start State Selection for Control Policy Learning from Optimal Trajectories

Christoph Zelch, O. Stryk

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Extended previous work by proposing a new supplementary strategy to generate starting points, improving the learning of approximate optimal state-dependent feedback control policies from optimal trajectories, and proving that learning only reaching near the target state is sufficient to achieve final stable control through PI control.

Statistical Safety and Robustness Guarantees for Feedback Motion Planning of Unknown Underactuated Stochastic Systems

Craig Knuth, Joseph L. Moore

Robotic Intelligence

🎯 What it does: Propose a method that provides statistical guarantees for runtime safety and goal reachability in the integrated planning and control of unknown nonlinear stochastic underactuated systems.

Statistical shape representations for temporal registration of plant components in 3D

K. Heiwolt, Grzegorz Cielniak

RecognitionRepresentation LearningTime SeriesAgriculture Related

🎯 What it does: Studied the use of leaf shape features to improve the matching of plant leaves in time series, and proposed an unlabeled shape compression algorithm to extract 3D shape features of leaves and achieve temporal association of leaves.

STD-Trees: Spatio-temporal Deformable Trees for Multirotors Kinodynamic Planning

Hongkai Ye, Fei Gao

OptimizationRobotic Intelligence

🎯 What it does: Proposes using space-time deformable trees to perform local optimization on trajectory trees for multirotor dynamic planning, aiming to improve convergence efficiency.

STD: Stable Triangle Descriptor for 3D place recognition

Chongjian Yuan, Fu Zhang

RecognitionAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposed the Stable Triangle Descriptor (STD) for 3D place recognition, and designed efficient keypoint extraction and triangle descriptor encoding methods.

Stealthy Perception-based Attacks on Unmanned Aerial Vehicles

Amir Khazraei, Miroslav Pajic

Anomaly DetectionAdversarial AttackRobotic IntelligenceImage

🎯 What it does: Study the vulnerability of UAVs under perception control to covert attacks, proposing methods for sustained attacks on sensor measurements and camera images in two tasks (ground vehicle tracking and vertical take-off/landing), leading to degraded control performance while maintaining stealth.

STEPS: Joint Self-supervised Nighttime Image Enhancement and Depth Estimation

Yupeng Zheng, Dong Zhao

RestorationDepth EstimationAutonomous DrivingImage

🎯 What it does: Proposed and implemented a self-supervised joint nighttime image enhancement and depth estimation framework without using any real labels.

StereoPose: Category-Level 6D Transparent Object Pose Estimation from Stereo Images via Back-View NOCS

Kai Chen, Qingxu Dou

Pose EstimationConvolutional Neural NetworkImage

🎯 What it does: Proposes StereoPose, a category-level 6D pose estimation framework for transparent objects based on stereo images, which includes steps such as object size estimation, initial pose estimation, and pose refinement.

StereoVAE: A lightweight stereo-matching system using embedded GPUs

Qiong Chang, Jun Miyazaki

Depth EstimationComputational EfficiencyAuto EncoderImageBenchmark

🎯 What it does: Proposes a lightweight stereo matching system using embedded GPU, which upsamples and refines the coarse disparity map generated by traditional matching methods through a variational autoencoder (VAE) network to achieve real-time processing.

StereoVoxelNet: Real-Time Obstacle Detection Based on Occupancy Voxels from a Stereo Camera Using Deep Neural Networks

Hongyu Li, T. Padır

Object DetectionDepth EstimationAutonomous DrivingConvolutional Neural NetworkImage

🎯 What it does: Propose a real-time obstacle detection method that directly detects occupied volumes from stereo images using deep neural networks, employing voxel representation and coarse-to-fine octree pruning based on decoder-generated octrees;

STEV: Stretchable Triboelectric E-skin enabled Proprioceptive Vibration Sensing for Soft Robot

Zihan Wang, Xiao-Ping Zhang

Robotic Intelligence

🎯 What it does: Proposed a stretchable electronic skin based on liquid metal and kirigami design to enable vibration sensing of soft robot bodies; and constructed a three-finger soft gripper with vibration sensing capabilities.

Stochastic Planning for ASV Navigation Using Satellite Images

Yizhou Huang, Florian Shkurti

Autonomous DrivingOptimizationImage

🎯 What it does: Use satellite images as a rough map to plan ASV sampling routes and propose a robust path planning algorithm to minimize the expected total travel distance.

Stochastic Robustness Interval for Motion Planning with Signal Temporal Logic

Roland B. Ilyes, Morteza Lahijanian

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Propose a new robustness metric for Signal Temporal Logic (STL) tailored to continuous-time stochastic trajectories, develop a monitor for partial trajectory inference based on this metric, and subsequently introduce a motion planning algorithm based on STL sampling for uncertain environments.

Stochastic Traveling Salesperson Problem with Neighborhoods for Object Detection

Cheng Peng, Volkan Isler

Object DetectionAutonomous DrivingOptimization

🎯 What it does: Proposes a new path planning problem considering perception and driving costs, modeled as a random traveling salesman problem with neighborhoods, and introduces a central visit method along with a 3D finite deviation scheme for non-overlapping neighborhoods.

STPOTR: Simultaneous Human Trajectory and Pose Prediction Using a Non-Autoregressive Transformer for Robot Follow-Ahead

Mohammad Mahdavian, Mo Chen

Pose EstimationRobotic IntelligenceTransformer

🎯 What it does: Propose a non-autoregressive Transformer model for simultaneously predicting human trajectory and pose to achieve robot front-following tasks.

Strained Elastic Surfaces with Adjustable-Modulus Edges (SESAMEs) for Soft Robotic Actuation

C. Kimmer, C. Harnett

Robotic Intelligence

🎯 What it does: Designed and implemented a soft robotic actuator that generates lightweight, resilient, and bendable torque through a stretchable elastic surface with adjustable modulus edges (SESAME).

Streaming LifeLong Learning With Any-Time Inference

S. Banerjee, Vinay P. Namboodiri

Computational EfficiencyKnowledge DistillationMeta Learning

🎯 What it does: Proposes a streaming lifelong learning method that supports single-sample single-pass learning, incremental class learning, and inference at any time; achieves fast parameter updates through a Bayesian framework, incorporates snapshot self-distillation for implicit regularization, and designs efficient online memory replay with sub-sample selection and novel buffer management strategies.

Structural Design and Frequency Tuning of Piezoelectric Energy Harvesters Based on Topology Optimization

Abbas Homayouni-Amlashi, Abdenbi Mohand-Ousaid

OptimizationPhysics Related

🎯 What it does: This paper optimizes the mechanical structure of a vibrational piezoelectric energy harvester using a topology optimization method.

Structure PLP-SLAM: Efficient Sparse Mapping and Localization using Point, Line and Plane for Monocular, RGB-D and Stereo Cameras

Fangwen Shu, D. Stricker

OptimizationSimultaneous Localization and MappingImage

🎯 What it does: Proposed a real-time visual SLAM system based on points, lines, and planes (PPR), achieving camera localization and sparse geometric reconstruction under multi-sensor conditions (monocular, RGB-D, stereo);

Structured Motion Generation with Predictive Learning: Proposing Subgoal for Long-Horizon Manipulation

Namiko Saito, S. Vijayakumar

GenerationRobotic Intelligence

🎯 What it does: Propose a prediction learning model based on deep neural networks with a Subgoal Proposal Module (SPM) to decompose long-horizon operations into short-horizon subgoals, evaluated on the long-horizon task of cutting and placing pizza.

Supernumerary Robotic Limbs for Next Generation Space Suit Technology

Erik Ballesteros, H. Asada

Robotic Intelligence

🎯 What it does: Developed and tested additional robotic arms (SuperLimbs) for NASA space suits to assist astronauts during extravehicular activities by grasping handrails and stabilizing their bodies.

Support Generation for Robot-Assisted 3D Printing with Curved Layers

Tianyu Zhang, Charlie C. L. Wang

Robotic Intelligence

🎯 What it does: Proposed a skeleton-based support generation method for surface layers in robot-assisted 3D printing.

Surgical-VQLA:Transformer with Gated Vision-Language Embedding for Visual Question Localized-Answering in Robotic Surgery

Long Bai, Hongliang Ren

Robotic IntelligenceTransformerVision Language ModelVideoMultimodality

🎯 What it does: Developed a visual question answering system for robotic-assisted surgery that can locate the relevant surgical area during the answering process.

Suture Thread Spline Reconstruction from Endoscopic Images for Robotic Surgery with Reliability-driven Keypoint Detection

Neelay Joglekar, Michael C. Yip

SegmentationRobotic IntelligenceBiomedical Data

🎯 What it does: Reconstructing 3D centerlines from segmented surgical image pairs using reliable keypoint detection and Minimum Variation Spline (MVS) smoothing optimization

Swarm Robotics Search and Rescue: A Bee-Inspired Swarm Cooperation Approach without Information Exchange

Yue Li, Q. Quan

Robotic Intelligence

🎯 What it does: Proposes a communication-free group collaboration method inspired by bee heuristics, including a target grouping approach for multi-object multi-robot scenarios, finite behavior state machines, and corresponding control laws, validated through simulation.

Swarm-LIO: Decentralized Swarm LiDAR-inertial Odometry

Fangcheng Zhu, Fu Zhang

Robotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a fully decentralized Swarm-LIO method for real-time LiDAR-inertial odometry in UAV swarms, encompassing self-state estimation, UAV detection and tracking, and relative state estimation.

SwinDepth: Unsupervised Depth Estimation using Monocular Sequences via Swin Transformer and Densely Cascaded Network

D. Shim, H. J. Kim

Depth EstimationTransformerVideo

🎯 What it does: This paper proposes an unsupervised monocular sequence depth estimation method, utilizing a convolution-free Swin Transformer as the feature extractor and designing a dense cascaded multi-scale network (DCMNet) for depth prediction.

Switching Attention in Time-Varying Environments via Bayesian Inference of Abstractions

Meghan Booker, Anirudha Majumdar

Robotic Intelligence

🎯 What it does: In time-varying environments, the study investigates how robots dynamically decide which environmental information to focus on, detect context changes, and switch attention mechanisms accordingly. It proposes a context-switching detection algorithm based on Bayesian inference and utilizes binary state abstraction to realize the switching of attention mechanisms.

Synthesizing Reactive Test Environments for Autonomous Systems: Testing Reach-Avoid Specifications with Multi-Commodity Flows

Apurva Badithela, R. Murray

OptimizationRobotic Intelligence

🎯 What it does: Using an LTL encoding system and test specifications, construct Büchi automata and specification product automata, generate a virtual product graph, and transform the test synthesis problem into a multi-commodity network flow optimization, thereby automatically generating constraints that satisfy the system and test specifications, forming a reactive testing environment.

Synthetic-to-Real Domain Adaptation for Action Recognition: A Dataset and Baseline Performances

Arun V. Reddy, Ramalingam Chellappa

RecognitionDomain AdaptationImageBenchmark

🎯 What it does: Created the RoCoG-v2 dataset and provided baseline experimental results

SyreaNet: A Physically Guided Underwater Image Enhancement Framework Integrating Synthetic and Real Images

J. Wen, Benwen Chen

RestorationData SynthesisDomain AdaptationImagePhysics Related

🎯 What it does: Proposed the SyreaNet framework, which integrates a synthesis module based on a revised model with a physics-guided decoupled network for seawater image enhancement;

System Configuration and Navigation of a Guide Dog Robot: Toward Animal Guide Dog-Level Guiding Work

Hochul Hwang, Donghyun Kim

Robotic Intelligence

🎯 What it does: A collaborative indoor navigation scheme was developed, combining speed and direction control, and utilizing semantic-aware local path planning to achieve safe and efficient guidance.

Tac-VGNN: A Voronoi Graph Neural Network for Pose-Based Tactile Servoing

Wen Fan, Dandan Zhang

Pose EstimationExplainability and InterpretabilityComputational EfficiencyRobotic IntelligenceGraph Neural Network

🎯 What it does: Developed a Voronoi diagram-based graph neural network (Tac-VGNN) for pose estimation and tactile servo control using the optical tactile sensor TacTip.

Tackling Clutter in Radar Data - Label Generation and Detection Using PointNet++

Johannes Kopp, K. Dietmayer

Object DetectionPoint Cloud

🎯 What it does: Developed two novel neural network architectures for clutter detection in radar data, proposed a method for automatically generating clutter labels, and constructed the first publicly available radar clutter dataset.

Tactile based robotic skills for cable routing operations

Andrea Monguzzi, P. Rocco

Robotic Intelligence

🎯 What it does: A set of tactile-based robotic skills for cable routing operations on deformable linear objects (DLOs) with significant stiffness and constrained at both ends, including reconstructing the shape of the gripped portion using tactile data, predicting future local shapes, aligning grasp configurations under rough initial grasping poses, and tracking the DLO's contour in three-dimensional space.

Tactile Identification of Object Shapes via In-Hand Manipulation with A Minimalistic Barometric Tactile Sensor Array

Xiaoxia Zhou, A. Spiers

RecognitionRobotic IntelligenceTime Series

🎯 What it does: Object Shape Recognition Using Three Adjacent Pressure Tactile Sensors Through Hand-Rolling

Tactile Tool Manipulation

Yuki Shirai, Dennis Hong

OptimizationRobotic Intelligence

🎯 What it does: Propose a closed-loop control method that uses tactile sensors to estimate the pose of tools and objects, and implements tool manipulation based on Model Predictive Control (MPC).

Tactile-Driven Gentle Grasping for Human-Robot Collaborative Tasks

Christopher J. Ford, N. Lepora

Robotic Intelligence

🎯 What it does: This paper develops a control scheme utilizing the Pisa/IIT SoftHand equipped with a miniature TacTip optical tactile sensor, achieving force-sensitive and gentle grasping through asynchronous sensor data acquisition and processing for real-time grasping control; subsequently, a new grasping controller is proposed, leveraging parallel feedback from five-finger tactile sensors to achieve gentle and stable grasping of 43 different geometric and stiffness objects, with this method applied to human-robot hand-over tasks.

TactoFind: A Tactile Only System for Object Retrieval

S. Pai, Pulkit Agrawal

RetrievalRobotic Intelligence

🎯 What it does: The study investigates a technical solution for object retrieval using tactile feedback in scenarios without visual information, unknown object shapes, and movable objects, i.e., locating, identifying, and grasping objects.

TANDEM3D: Active Tactile Exploration for 3D Object Recognition

Jingxi Xu, M. Ciocarlie

RecognitionRobotic IntelligenceConvolutional Neural NetworkPoint Cloud

🎯 What it does: Proposes a method called TANDEM3D, which utilizes a co-training framework for tactile exploration and decision-making to achieve 3D object recognition. The method is based on a novel encoder derived from PointNet++, which constructs 3D object representations from contact points and normals, and employs 6DOF motion to enable efficient tactile information collection. The entire model is trained in a simulated environment and validated through real-world experiments.

Target-Aware Implicit Mapping for Agricultural Crop Inspection

Shane P. Kelly, C. Stachniss

Robotic IntelligenceSimultaneous Localization and MappingImageAgriculture Related

🎯 What it does: A dense 3D map of strawberry and sweet pepper crop rows with high precision was constructed using RGB images captured by a wheeled mobile field robot in a greenhouse, leveraging implicit mapping technology to document the growth status of plants and fruits.