IROS 2023 Papers — Page 10
IEEE/RSJ International Conference on Intelligent Robots and Systems · 1195 papers
Real-Time Tube-Based Non-Gaussian Risk Bounded Motion Planning for Stochastic Nonlinear Systems in Uncertain Environments via Motion Primitives
Weiqiao Han, B. Williams
Autonomous Driving
🎯 What it does: Propose a real-time online motion planning algorithm for long-term tasks of stochastic nonlinear systems in uncertain environments, utilizing offline-constructed discrete-time motion primitives and corresponding continuous-time tubes, and verifying safety through SOS programming during the online phase.
Real-Time Video Inpainting for RGB-D Pipeline Reconstruction
Luyuan Wang, Lu Li
RestorationDepth EstimationComputational EfficiencyConvolutional Neural NetworkSimultaneous Localization and MappingOptical FlowVideoPoint Cloud
🎯 What it does: Proposed a real-time video inpainting algorithm that enables a single-camera-laser pipeline inspection robot to simultaneously acquire color and 3D information through a single video stream.
Real-Time Whole-Body Collision Avoidance and Path Following of a Snake Robot Through MPC-based Optimization Strategies
Liuyin Wang, Yantao Shen
OptimizationRobotic Intelligence
🎯 What it does: A real-time optimized whole-body collision avoidance and path tracking strategy is proposed using model predictive control (MPC), verified on a nine-segment snake robot.
Real2Sim2Real Transfer for Control of Cable-Driven Robots Via a Differentiable Physics Engine
Kun Wang, Kostas E. Bekris
Domain AdaptationRobotic IntelligenceWorld ModelTabularTime SeriesPhysics Related
🎯 What it does: Proposed a Real2Sim2Real (R2S2R) strategy that trains on limited real robot data using a differentiable physics engine to generate locomotion policies directly transferable to real 3-bar tensegrity robots.
Recognizing Real-World Intentions using A Multimodal Deep Learning Approach with Spatial-Temporal Graph Convolutional Networks
Jiaqi Shi, Hiroshi Ishiguro
ClassificationRecognitionGraph Neural NetworkImageMultimodalityGraph
🎯 What it does: Collect real-time behavioral data at building entrances using hand part allocators and temperature scanners, and classify intentions through skeleton data and image features; propose a skeleton intention recognition method based on a spatial-temporal graph convolutional network, as well as a framework for automatically inferring intentions using multi-modal deep learning.
Recurrent Macro Actions Generator for POMDP Planning
Yuanchu Liang, Hanna Kurniawati
OptimizationRecurrent Neural NetworkReinforcement Learning
🎯 What it does: This paper proposes a simple recurrent neural network for generating appropriate macro action sets in POMDP planning, thereby reducing the effective planning cycle and lowering computational complexity.
Reducing Safety Interventions in Provably Safe Reinforcement Learning
Jakob Thumm, M. Althoff
Reinforcement LearningBenchmark
🎯 What it does: Investigated two methods for reducing safety interventions and evaluated them on the OpenAI Safety Gym benchmark and human-robot collaboration tasks
Reducing Workload During Brain Surgery with Robot-Assisted Autonomous Exoscope
Elisa Iovene, E. Momi
Robotic IntelligenceBiomedical Data
🎯 What it does: Developed and verified a position-based visual servo control method for a robotic camera mount to improve ergonomics and reduce surgeons' workload during neurosurgical procedures.
Refining 6-DoF Grasps with Context-Specific Classifiers
Tasbolat Taunyazov, Harold Soh
Pose EstimationRobotic IntelligenceGenerative Adversarial Network
🎯 What it does: Proposes the GraspFlow framework, which uses the discriminator gradient flow method to perform context-specific refinement of 6-DoF grasp poses;
Reinforced Potential Field for Multi-Robot Motion Planning in Cluttered Environments
Dengyu Zhang, Qingrui Zhang
OptimizationRobotic IntelligenceTransformerReinforcement Learning
🎯 What it does: This paper proposes a reinforced potential field method for distributed multi-robot motion planning in crowded environments, integrating reinforcement learning with artificial potential fields;
Reinforcement Learning Based Multi-Layer Bayesian Control for Snake Robots in Cluttered Scenes
Jessica Ziyu Qu, Yuanyuan Jia
Robotic IntelligenceRecurrent Neural NetworkReinforcement Learning
🎯 What it does: Proposed a multi-layer Bayesian method based on reinforcement learning for autonomous snake robot control, aiming to address spatial and temporal dependencies within the robot and its interactions with the environment.
Reinforcement Learning for Robot Navigation with Adaptive Forward Simulation Time (AFST) in a Semi-Markov Model
Yu'an Chen, Yanyong Zhang
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a robot navigation method AFST based on deep reinforcement learning, using semi-Markov decision process (SMDP) and continuous action space to address the local minima problem.
Reinforcement Learning Under Probabilistic Spatio-Temporal Constraints with Time Windows
Xiaoshan Lin, Derya Aksaray
OptimizationReinforcement Learning
🎯 What it does: Propose a reinforcement learning method based on automata to address decision-making problems under complex spatiotemporal constraints with time windows
Relationship Between Ankle Assistive Torque and Biomechanical Gait Metrics in Individuals After Stroke
Jesús de Miguel-Fernández, J. Lobo-Prat
Biomedical Data
🎯 What it does: Using the wearable ankle exoskeleton (ABLE-S) on five stroke survivors, altering the peak torque and timing of foot dorsiflexion/plantarflexion, recording gait biomechanical metrics, and analyzing their correlation with exoskeleton control parameters.
Relative Roughness Measurement Based Real-Time Speed Planning for Autonomous Vehicles on Rugged Road
Liang Wang, Junzheng Wang
Autonomous DrivingOptimizationPoint Cloud
🎯 What it does: Designed a real-time speed planning method based on the time-frequency domain transformation of LiDAR point cloud vertical profiles to achieve adaptive speed control on rough roads.
Residual Physics Learning and System Identification for Sim-to-real Transfer of Policies on Buoyancy Assisted Legged Robots
Nitish Sontakke, Sehoon Ha
Robotic IntelligenceReinforcement LearningPhysics Related
🎯 What it does: Robust simulation-to-real strategy transfer on the BALLU robot was achieved through system identification and a novel residual physical learning method (EnvMimic).
Resilient and Distributed Multi-Robot Visual SLAM: Datasets, Experiments, and Lessons Learned
Yulun Tian, L. Carlone
Robotic IntelligenceSimultaneous Localization and MappingBenchmark
🎯 What it does: Improve Kimera-Multi to enhance robustness in large-scale real-world environments, collect and release the MIT campus field dataset containing multi-robot, multi-distance, indoor-outdoor hybrid, visual ambiguity, and dynamic entities, evaluate its performance under different communication scenarios, and quantitatively compare it with centralized baseline systems.
Resource-Constrained Station-Keeping for Latex Balloons Using Reinforcement Learning
Jack D. Saunders, Wenbin Li
Reinforcement LearningPhysics Related
🎯 What it does: Continuous control of a low-cost high-altitude balloon equipped with air pump venting and ballast is achieved using reinforcement learning (soft actor-critic algorithm) to maintain station-keeping within the stratosphere, with physical feasibility of actions ensured through motion equation constraints.
Revisiting Deformable Convolution for Depth Completion
Xinglong Sun, Yu-Xiong Wang
Depth EstimationAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: Propose a single inference module utilizing deformable convolutions for generating high-quality dense depth maps from sparse depth maps.
Revisiting Event-Based Video Frame Interpolation
Jiaben Chen, Shenghua Gao
GenerationOptical FlowVideo
🎯 What it does: Propose to utilize RGB information to guide event optical flow refinement, and decompose event-driven video frame interpolation into multi-stage incremental synthesis to fully exploit the high temporal density and noise characteristics of events.
Reward Shaping for Building Trustworthy Robots in Sequential Human-Robot Interaction
Yaohui Guo, Cong Shi
Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: Model the human-robot interaction (HRI) process as a two-player Markov game, and use reward shaping techniques to enhance human trust while limiting task performance loss; under this framework, convert the experience-based trust model into linear programming, enabling efficient solution and deployment in practical applications; validate the framework in a search and rescue simulation scenario, demonstrating that it can improve human trust under robot strategies with minimal task cost.
RFDNet: Real-Time 3D Object Detection Via Range Feature Decoration
Hongda Chang, Jun Cheng
Object DetectionAutonomous DrivingComputational EfficiencyPoint Cloud
🎯 What it does: Proposed a real-time 3D object detection framework named RFDNet
RGBD Fusion Grasp Network with Large-Scale Tableware Grasp Dataset
J. Yoon, Sungchul Kang
Pose EstimationRobotic IntelligenceMultimodality
🎯 What it does: Propose an RGBD fusion-based grasping network, construct a large-scale utensil grasping dataset, and design a stable grasping pose sampling method.
Risk-Aware Emergency Landing Planning for Gliding Aircraft Model in Urban Environments
Jakub Sláma, J. Faigl
Optimization
🎯 What it does: Planning a safe emergency landing path for aircraft when thrust is lost, particularly assessing risks and identifying the safest landing location and trajectory when no airport is reachable.
Risk-Aware Safe Control for Decentralized Multi-Agent Systems via Dynamic Responsibility Allocation
Yiwei Lyu, J. Dolan
🎯 What it does: A risk-aware decentralized multi-agent control framework is proposed, which dynamically allocates responsibility shares through risk measurement inspired by control barrier functions (CBF), achieving collision avoidance and efficient movement without direct communication.
Risk-Aware Stochastic Ship Routing Using Conditional Value-at-Risk
Andre Nuñez, Robert Fitch
OptimizationTime Series
🎯 What it does: Investigated the use of Conditional Value at Risk (CVaR) as an objective and constraint function in ship path planning to balance safety and efficiency.
Risk-Sensitive Mobile Robot Navigation in Crowded Environment via Offline Reinforcement Learning
Jiaxu Wu, A. Yamashita
Autonomous DrivingRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a multi-strategy control framework that integrates offline reinforcement learning navigation policies with risk detectors and force-based risk avoidance strategies to achieve risk-sensitive mobile robot navigation in crowded environments.
Risk-Tolerant Task Allocation and Scheduling in Heterogeneous Multi-Robot Teams
Jinwoo Park, Seth Hutchinson
OptimizationRobotic Intelligence
🎯 What it does: Developed a risk-tolerant task allocation and scheduling algorithm that employs the Sequential Probability Ratio Test (SPRT) and Mixed-Integer Linear Programming (MILP) to handle capability and task uncertainty in heterogeneous multi-robot teams, while optimizing under time deadlines, synchronization, and precedence constraints;
RLPG: Reinforcement Learning Approach for Dynamic Intra-Platoon Gap Adaptation for Highway On-Ramp Merging
Sushma Reddy Yadavalli, M. Won
Reinforcement Learning
🎯 What it does: Designed, implemented, and evaluated a reinforcement learning framework for dynamically adjusting vehicle spacing within a platoon during highway on-ramp merging processes to maximize traffic flow.
RMBench: Benchmarking Deep Reinforcement Learning for Robotic Manipulator Control
Yanfei Xiang, Siwei Lyu
Robotic IntelligenceReinforcement LearningImageBenchmark
🎯 What it does: Proposed RMBench, the first benchmark for robotic manipulation in high-dimensional continuous action and state spaces; implemented and evaluated deep reinforcement learning algorithms using pixel inputs, reporting their average performance and learning curves.
Roblets: Robotic Tablets That Self-Assemble and Self-Fold into a Robot
Junyi Han, S. Miyashita
Robotic Intelligence
🎯 What it does: Self-assembling robots: Random parts are first assembled into a 2D structure, then self-fold into a 3D shape via heat-responsive thin films, resulting in robots containing magnets that can complete basic tasks such as pushing blocks under external magnetic field actuation.
Robo-Centric ESDF: A Fast and Accurate Whole-Body Collision Evaluation Tool for Any-Shape Robotic Planning
Shuang Geng, Fei Gao
OptimizationRobotic Intelligence
🎯 What it does: Proposes a robot-centered ESDF (RC-ESDF) for whole-body collision evaluation of arbitrary-shaped mobile robots, and jointly optimizes robot position and rotation while considering whole-body safety, smoothness, and dynamic feasibility;
Robot Learning to Mop Like Humans Using Video Demonstrations
S. Gaurav, Brian D. Ziebart
Robotic IntelligenceReinforcement Learning from Human FeedbackConvolutional Neural NetworkReinforcement LearningVideo
🎯 What it does: Developed a robotic system that mimics human floor mopping behavior from video demonstrations.
Robot Team Data Collection with Anywhere Communication
Matthew A. Schack, Neil T. Dantam
OptimizationRobotic Intelligence
🎯 What it does: Propose a mixed-integer linear programming (MILP) model for robot path planning to reduce data collection delay.
Robot-Induced Group Conversation Dynamics: A Model to Balance Participation and Unify Communities
Lucrezia Grassi, A. Sgorbissa
Robotic Intelligence
🎯 What it does: Studied the impact of robots in group conversations and evaluated the effectiveness of different addressing strategies.
Robotic Barrier Construction through Weaved, Inflatable Tubes
Heather Jin Hee Kim, H. Kao
Robotic Intelligence
🎯 What it does: Proposed a lightweight obstacle construction mechanism based on expandable tubes and developed a corresponding path planning algorithm to integrate the tubes into existing environmental features.
Robotic Crop Handling in Cluttered and Unstructured Environments using Simulated L-System Dynamic Plant Models
Quinlan T. Barthelme, Chris Lehnert
OptimizationRobotic IntelligenceWorld ModelAgriculture Related
🎯 What it does: Developed an L-system-based dynamic plant model simulation to evaluate and optimize robot methods for handling crops in cluttered environments.
Robotic Defect Inspection with Visual and Tactile Perception for Large-Scale Components
Arpit Agarwal, Wenzhen Yuan
Anomaly DetectionRobotic IntelligenceImageMultimodality
🎯 What it does: Proposed a two-stage multi-modal detection pipeline based on visual and tactile perception for defect detection and localization in large-scale parts.
Robotic Kinematic Calibration with Only Position Data and Consideration of Non-Geometric Errors Using POE-Based Model and Gaussian Mixture Models
Xiao Luo, Zheng Li
Robotic Intelligence
🎯 What it does: Proposes a robot kinematic calibration algorithm based on an enhanced POE model and Gaussian Mixture Models (GMMs), which can consider and compensate for non-geometric errors that traditional models cannot fit using only position data.
Robotic Powder Grinding with Audio-Visual Feedback for Laboratory Automation in Materials Science
Yusaku Nakajima, K. Ono
Robotic IntelligenceImageMultimodalityAudio
🎯 What it does: Developed a multimodal robotic powder grinding system that utilizes audio and visual feedback.
Robotic Scene Segmentation with Memory Network for Runtime Surgical Context Inference
Zongyu Li, H. Alemzadeh
SegmentationRobotic IntelligenceVideoBiomedical Data
🎯 What it does: Propose a Space-Time Correspondence Memory Network (STCN) for real-time scene segmentation and context inference in robotic surgery.
Robots as AI Double Agents: Privacy in Motion Planning
Rahul Shome, L. Kavraki
Safty and PrivacyRobotic Intelligence
🎯 What it does: This paper proposes and discusses the potential privacy invasion risks posed by robots in autonomous motion planning, and demonstrates through simulation cases how privacy leakage can be easily achieved by adjusting the cost function;
Robottheory Fitness: GoBot's Engagement Edge for Spurring Physical Activity in Young Children
Rafael Morales Mayoral, Naomi T. Fitter
Robotic Intelligence
🎯 What it does: In a two-month experiment with weekly sessions, 8 children interacted with a custom GoBot robot, testing three modes: remote operation, semi-autonomous, and control (robot inactive), while recording children's activity levels.
Robust Electric Vehicle Balancing of Autonomous Mobility-on-Demand System: A Multi-Agent Reinforcement Learning Approach
Sihong He, Fei Miao
Autonomous DrivingOptimizationReinforcement Learning
🎯 What it does: Propose a multi-agent reinforcement learning framework and design a robust E-AMoD balancing algorithm to address the uncertainties in electric vehicle supply and demand.
Robust Fusion for Bayesian Semantic Mapping
David Morilla-Cabello, E. Montijano
SegmentationAutonomous DrivingRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Propose a robust fusion method that combines Bayesian semantic prediction, reducing overconfident anomaly predictions through uncertainty calibration in the fusion process. The method regularizes and weights observations using uncertainty estimation from Bayesian neural networks, forming a new probabilistic distribution to achieve more reliable integration of semantic information in maps.
Robust Generalized Proportional Integral Control for Trajectory Tracking of Soft Actuators in a Pediatric Wearable Assistive Device
Caio Mucchiani, Konstantinos Karydis
Robotic IntelligenceTime Series
🎯 What it does: Developed a robust generalized proportional-integral controller for trajectory tracking of soft pneumatic actuators in children's wearable assistive devices;
Robust Localization of Aerial Vehicles via Active Control of Identical Ground Vehicles
Igor Spasojevic, Vijay Kumar
Autonomous DrivingOptimization
🎯 What it does: Designed and verified an active cooperative localization method by controlling homogeneous UGVs at predetermined positions, leveraging unlabeled measurements from UGVs to UAVs to enable UAVs to uniquely determine their global pose.
Robust Point Cloud Registration with Geometry-based Transformation Invariant Descriptor
Jianjie Lin, Alois Knoll
Pose EstimationRepresentation LearningGraph Neural NetworkPoint Cloud
🎯 What it does: A robust point cloud registration method based on geometric invariant transformation descriptors is proposed. Three geometric descriptors are used to filter pseudo-correspondences, and the problem of outlier removal is transformed into a subgraph isomorphism problem by constructing a fully connected graph. Finally, a binary clustering approach is employed to obtain the inlier set for transformation estimation.
Robust Real-Time Motion Retargeting via Neural Latent Prediction
Tiantian Wang, Rong Xiong
Pose EstimationComputational EfficiencyGraph Neural Network
🎯 What it does: Proposes a robust real-time motion retargeting method based on neural latent prediction, combining spatiotemporal graph structure-based motion retargeting and latent space motion prediction, utilizing a prediction-oriented controller to compensate for computational delays, enhancing synchronization and similarity while maintaining fault tolerance.
Robust Satisfaction of Joint Position and Velocity Bounds in Discrete-Time Acceleration Control of Robot Manipulators
Erik Zanolli, Andrea Del Prete
OptimizationRobotic Intelligence
🎯 What it does: Developed a method utilizing mathematical and computational tools to ensure that the fully actuated robotic system meets joint position, velocity, and acceleration bounds under constrained disturbances.
Robust Self-Supervised Extrinsic Self-Calibration
Takayuki Kanai, Rares Ambrus
Pose EstimationDepth EstimationAutonomous DrivingImageBenchmark
🎯 What it does: Proposes a multi-camera extrinsic self-calibration method based on self-supervised monocular depth estimation and pose learning.
Robust Unmanned Surface Vehicle Navigation with Distributional Reinforcement Learning
Xi Lin, Brendan Englot
Autonomous DrivingReinforcement Learning
🎯 What it does: Propose a local path planner based on distributed reinforcement learning (DRL), which learns return distributions and adaptively adjusts risk sensitivity to address the perceptual navigation problem of unmanned surface vessels in unknown ocean currents and obstacle environments.
Robust Visual Sim-to-Real Transfer for Robotic Manipulation
Ricardo Garcia Pinel, C. Schmid
Domain AdaptationRobotic IntelligenceReinforcement LearningImage
🎯 What it does: Learn a visual-motor strategy using domain randomization in a simulated environment, and optimize the randomization parameters through an offline proxy task (cube positioning), ultimately achieving efficient simulation-to-reality transfer on a real robot.
Roller-Quadrotor: A Novel Hybrid Terrestrial/Aerial Quadrotor with Unicycle-Driven and Rotor-Assisted Turning
Zhi Zheng, Fei Gao
Robotic Intelligence
🎯 What it does: This paper designs, models, and experimentally verifies a hybrid land-air quadrotor robot named Roller-Quadrotor.
Rollvox: Real-Time and High-Quality LiDAR Colorization with Rolling Shutter Camera
Sheng Hong, Shaojie Shen
Autonomous DrivingSimultaneous Localization and MappingOptical FlowImagePoint Cloud
🎯 What it does: This study proposes a system that uses a low-cost rolling shutter camera to perform real-time high-quality coloring of LiDAR point clouds.
Rotating Objects via in-Hand Pivoting Using Vision, Force and Touch
Shiyu Xu, Akansel Cosgun
Robotic IntelligenceMultimodality
🎯 What it does: Achieve rotation of held objects around the grasp point using visual, wrist force, and tactile sensing, allowing rotational slip while preventing translational slip.
RREx-BoT: Remote Referring Expressions with a Bag of Tricks
Gunnar A. Sigurdsson, Robinson Piramuthu
Object DetectionVision Language ModelMultimodalityAgriculture Related
🎯 What it does: Use a general vision-language scoring model (with minor modifications to 3D encoding) for remote object localization in observed environments.
Run and Catch: Dynamic Object-Catching of Quadrupedal Robots
Yangwei You, Shiwu Zhang
Robotic IntelligenceVideo
🎯 What it does: Developed a stereo vision-based control pipeline enabling quadruped robots to dynamically capture flying balls while in motion.
RVWO: A Robust Visual-Wheel SLAM System for Mobile Robots in Dynamic Environments
J. Mahmoud, S. Kolyubin
Robotic IntelligenceSimultaneous Localization and MappingImage
🎯 What it does: Proposed a robust visual-wheel SLAM system (RVWO) for localization and mapping of mobile robots in dynamic environments
Safe Active Learning and Probabilistic Design of Experiment for Autonomous Hydraulic Excavators
Maximilian Dio, Knut Graichen
Robotic Intelligence
🎯 What it does: Proposed and implemented two methods, static learning and active learning, to reduce the amount of data required to learn a hydraulic inverse execution model.
Safe Collision and Clamping Reaction for Parallel Robots During Human-Robot Collaboration
Aran Mohammad, T. Ortmaier
Safty and PrivacyRobotic Intelligence
🎯 What it does: Studied and implemented safety response strategies for collision and grasping scenarios during human-robot collaboration on planar parallel robots, including external force estimation, immediate retreat, damping reduction, and grasping chain classification and structural opening based on neural networks; validated its effectiveness through real experiments.
Safety-Assured Speculative Planning with Adaptive Prediction
Xiangguo Liu, Qi Zhu
Autonomous DrivingOptimizationReinforcement Learning
🎯 What it does: Proposes a safety-guaranteed speculative planning framework that quantifies uncertainty at the behavioral and trajectory levels of surrounding vehicles through a prediction-planning interface, maximizing the ego vehicle's expected reward while excluding potentially unsafe actions under worst-case scenarios.
Safety-Critical Coordination for Cooperative Legged Locomotion via Control Barrier Functions
Jeeseop Kim, A. Ames
Robotic Intelligence
🎯 What it does: Proposes a safety-critical cooperative control method based on control barrier functions (CBF), ensuring safety, formation maintenance, and obstacle avoidance for multiple robots walking in fixed constrained environments.
Sample-Efficient Real-Time Planning with Curiosity Cross-Entropy Method and Contrastive Learning
Mostafa Kotb, Stefan Wermter
Reinforcement LearningContrastive LearningImage
🎯 What it does: Proposed the Curiosity CEM (CCEM) algorithm, which improves CEM by incorporating curiosity-driven exploration in model-based reinforcement learning to enhance planning efficiency in high-dimensional environments.
SBlimp: Design, Model, and Translational Motion Control for a Swing-Blimp
Jiawei Xu, D. Saldaña
Robotic IntelligencePhysics Related
🎯 What it does: Designed and built an aerial vehicle called SBlimp, which combines a tilted-rotor quadrotor with a helium balloon, and proposed a control strategy that utilizes the balloon's self-stabilizing attitude to achieve translational motion.
Scalable. Intuitive Human to Robot Skill Transfer with Wearable Human Machine Interfaces: On Complex, Dexterous Tasks
Felipe Sanches, Minas V. Liarokapis
Robotic IntelligenceOptical Flow
🎯 What it does: Designed and verified an efficient human-robot skill transfer method based on wearable interfaces and optical tracking, enabling robot arm-hand systems to perform complex dexterous tasks in dynamic environments.
Scale Jump-Aware Pose Graph Relaxation for Monocular SLAM with Re-Initializations
Runze Yuan, L. Kneip
Robotic IntelligenceSimultaneous Localization and MappingImage
🎯 What it does: Propose a hybrid pose graph relaxation method that can recover globally consistent trajectories under monocular SLAM reinitialization and unknown scale between subsequent frames, particularly suitable for small indoor service robots capable only of pure rotational displacement.
Scaling Vision-Based End-to-End Autonomous Driving with Multi-View Attention Learning
Yi Xiao, Antonio M. López
Autonomous DrivingTransformerImage
🎯 What it does: Propose the CIL++ model, improving upon CILRS by utilizing higher-resolution images and human-inspired HFOV prior, while incorporating a multi-view attention mechanism for end-to-end training;
ScAR: Scaling Adversarial Robustness for LiDAR Object Detection
Xiaohu Lu, H. Radha
Object DetectionAutonomous DrivingAdversarial AttackPoint Cloud
🎯 What it does: Proposed the ScAR method, achieving improvements in black-box scaling adversarial robustness for LiDAR object detection.
SCRIMP: Scalable Communication for Reinforcement- and Imitation-Learning-Based Multi-Agent Pathfinding
Yutong Wang, Guillaume Sartoretti
OptimizationTransformerReinforcement Learning
🎯 What it does: Proposes the SCRIMP framework, enabling multi-agent systems to learn collaborative path planning strategies through global communication within a very small 3x3 field of view (FOV).
SCTOMP: Spatially Constrained Time-Optimal Motion Planning
Jon Arrizabalaga, Markus Ryll
Optimization
🎯 What it does: Proposes a three-stage spatiotemporal optimal motion planning method that can generate dynamically constrained optimal trajectories based solely on start, end, and environmental geometry information.
SDF-Pack: Towards Compact Bin Packing with Signed-Distance-Field Minimization
Jia-Yu Pan, Chi-Wing Fu
OptimizationRobotic Intelligence
🎯 What it does: Proposes the SDF-Pack method based on signed distance fields (SDF), representing object geometry with truncated SDF, and computes object placement positions and arrangement sequences within containers through an SDF-minimization heuristic algorithm to achieve more compact robotic bin packing.
SDFMAP: Neural Signed Distance Fields for Mapping and Positioning in Real-Time
Shaofan Liu, Jianke Zhu
Pose EstimationNeural Radiance FieldSimultaneous Localization and MappingImage
🎯 What it does: Propose an end-to-end neural network called SDFMAP, which can perform camera pose estimation and indoor scene reconstruction in real-time conditions by learning a truncated signed distance function (TSDF).
SEAL: Simultaneous Exploration and Localization for Multi-Robot Systems
Ehsan Latif, Ramviyas Parasuraman
OptimizationRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Propose the SEAL method to achieve simultaneous exploration and localization for multiple robots, and maximize exploration and relative localization through Gaussian process information fusion and communication graph optimization.
SeasonDepth: Cross-Season Monocular Depth Prediction Dataset and Benchmark Under Multiple Environments
Hanjiang Hu, Hesheng Wang
Depth EstimationImageBenchmark
🎯 What it does: Proposed the cross-season monocular depth prediction dataset SeasonDepth, conducted benchmarking of depth estimation performance under different environments, and evaluated multiple advanced supervised and self-supervised depth prediction methods.
See What a Strabismus Patient Sees Using Eye Robots
Yidi Huang, Ningshi Yao
Robotic IntelligenceImageBiomedical Data
🎯 What it does: Developed the first set of visual visualization robots that simulate the visual scenes of strabismus patients using eye movement data, constructed a binocular eye platform, and achieved time-varying perspective fusion based on homotopy transformation;
See What the Robot Can't See: Learning Cooperative Perception for Visual Navigation
Jan Blumenkamp, Amanda Prorok
Robotic IntelligenceGraph Neural NetworkImage
🎯 What it does: Train sensor encoders and communicate with robots to achieve visual navigation without localization;
Seeing the Fruit for the Leaves: Robotically Mapping Apple Fruitlets in a Commercial Orchard
A. Qureshi, Henry Williams
Robotic IntelligenceSimultaneous Localization and MappingImageAgriculture Related
🎯 What it does: Developed a visual system for an automated apple fruit pruning robot, capable of precisely measuring and mapping fruits on each tree in commercial orchards.
Selective Presentation of AI Object Detection Results While Maintaining Human Reliance
Yosuke Fukuchi, Seiji Yamada
Object DetectionAutonomous Driving
🎯 What it does: Propose the SmartBBox method, which intelligently selectively displays AI detection result bounding boxes to reduce information overload
Self-Supervised Drivable Area Segmentation Using LiDAR's Depth Information for Autonomous Driving
Fulong Ma, Ming Liu
SegmentationAutonomous DrivingPoint Cloud
🎯 What it does: This paper proposes an automatic data annotator (ADL) that utilizes LiDAR depth information to achieve ground plane segmentation and road boundary detection, and trains a semantic segmentation network based on this.
Self-Supervised Event-Based Monocular Depth Estimation Using Cross-Modal Consistency
Junyu Zhu, Yong Liu
Depth EstimationConvolutional Neural NetworkContrastive LearningMultimodality
🎯 What it does: Proposed a self-supervised event camera monocular depth estimation framework called EMoDepth.
Self-Supervised Instance Segmentation by Grasping
Yuxuan Liu, P. Abbeel
SegmentationData SynthesisRobotic IntelligenceImage
🎯 What it does: Acquire self-supervised instance segmentation labels through grasping interactions, train a grasping segmentation model, generate instance segmentation data using the cut-and-paste method, thereby enhancing instance segmentation performance and reducing grasping error rates.
Self-Supervised Object Goal Navigation with In-Situ Finetuning
So Yeon Min, Jian Zhang
Domain AdaptationRobotic IntelligenceSupervised Fine-TuningContrastive LearningImage
🎯 What it does: A goal-object navigation agent based on self-supervised learning was constructed, trained using location consistency (LocCon) in unlabeled simulated houses, and achieved real-world self-supervised fine-tuning;
Self-Supervised Visual Motor Skills via Neural Radiance Fields
Paul Gesel, M. Begum
Robotic IntelligenceNeural Radiance Field
🎯 What it does: Designed a self-supervised visual motion policy network architecture that utilizes NeRF and keypoint correspondence to achieve visual imitation learning.
SELVO: A Semantic-Enhanced Lidar-Visual Odometry
Kun Jiang, Jianyu Wang
Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingImagePoint Cloud
🎯 What it does: Proposed a semantic-enhanced LiDAR-visual odometry (SELVO) that achieves high-precision pose estimation by fusing camera and LiDAR information and leveraging semantic information.
Semantic Scene Difference Detection in Daily Life Patroling by Mobile Robots Using Pre-Trained Large-Scale Vision-Language Model
Yoshiki Obinata, M. Inaba
Robotic IntelligenceLarge Language ModelImageText
🎯 What it does: This paper proposes a Visual Question Answering (VQA) method based on a large vision-language model, which detects semantic changes in daily environments by comparing multiple questions from reference images and current images to obtain sentence answers, and verifies it through patrol tasks on the Fetch Mobile Manipulator mobile robot.
Semantic Segmentation Based on Multiple Granularity Learning
Kebin Wu, M. Debbah
SegmentationImage
🎯 What it does: Proposes a semantic segmentation algorithm based on multi-granularity learning, SSMGL, which enhances segmentation performance through regularized representation space.
Semantically Informed MPC for Context-Aware Robot Exploration
∗. YashGoel, C. Stachniss
Robotic IntelligenceImage
🎯 What it does: This paper studies goal object navigation tasks in unknown environments with semantic labels, proposing a deep neural network framework based on continuous control. It uses information-theoretic model predictive control (MPC) on dense cost maps to guide robots toward target objects, while integrating mid-level visual representations to provide additional semantic cues.
Semantically-Enhanced Deep Collision Prediction for Autonomous Navigation Using Aerial Robots
M. Kulkarni, K. Alexis
Robotic IntelligenceAuto Encoder
🎯 What it does: Proposed a modular learning approach for navigation in crowded environments.
SemanticBEVFusion: Rethinking LiDAR-Camera Fusion in Unified Bird's-Eye View Representation for 3D Object Detection
Qi Jiang, Hao Sun
Object DetectionAutonomous DrivingImageMultimodalityPoint Cloud
🎯 What it does: Proposes SemanticBEVFusion, which deeply integrates camera features and LiDAR features within a unified bird's-eye view (BEV) representation to enhance 3D object detection performance.
Semantics-Aware Mission Adaptation for Autonomous Exploration in Urban Environments
Sangwoo Moon, A. Agha-mohammadi
Robotic Intelligence
🎯 What it does: Proposes a task planning adaptation framework that utilizes real-time semantic information for autonomous exploration and wireless source tracking in multi-story building environments, with its effectiveness validated through simulation and real-world experiments.
Semi-Autonomous Assistance for Telesurgery Under Communication Loss
Hisashi Ishida, P. Kazanzides
Robotic IntelligenceBiomedical Data
🎯 What it does: Proposed a remote surgery simulation framework to analyze personnel responses during communication loss and compared user-centered, robot-centered, and hybrid assistance modes in a peg transfer task involving 12 participants.
Sensor Selection for Fine-Grained Behavior Verification that Respects Privacy
Rishi Phatak, Dylan A. Shell
OptimizationSafty and Privacy
🎯 What it does: Proposes a sensor selection method under multi-behavior routes, balancing behavior verification and privacy protection.
Sequential Manipulation Planning for Over-Actuated Unmanned Aerial Manipulators
Yao Su, Hangxin Liu
Robotic Intelligence
🎯 What it does: Studied the sequential manipulation planning problem for unmanned aerial manipulators (UAM), using the Virtual Kinematic Chain (VKC) framework to uniformly model and plan the cooperative motion of the floating base, manipulator arm, and manipulated object.
Sequential Neural Barriers for Scalable Dynamic Obstacle Avoidance
Hong-Den Yu, Sicun Gao
Autonomous DrivingSequential
🎯 What it does: Designed an scalable dynamic obstacle avoidance method, utilizing Sequential Neural Control Barrier Functions (SN-CBFs) to achieve safe navigation among multiple dynamic obstacles.
SG-LSTM: Social Group LSTM for Robot Navigation Through Dense Crowds
Rashmi Bhaskara, Aniket Bera
Autonomous DrivingRobotic IntelligenceRecurrent Neural NetworkVideo
🎯 What it does: Proposed the SG-LSTM model for predicting pedestrian movement and interactions in crowded environments, enhancing robot navigation safety and efficiency in complex crowds.
Shape Completion with Prediction of Uncertain Regions
Matthias Humt, U. Hillenbrand
RestorationGenerationImage
🎯 What it does: Propose two shape completion methods for predicting the shape of uncertain regions, and generate and use a real rendered depth image dataset derived from ShapeNet for training and evaluation.
Shape Control of Variable Length Continuum Robots Using Clothoid-Based Visual Servoing
Abhinav Gandhi, B. Çalli
Robotic Intelligence
🎯 What it does: Proposes a visual servoing method based on clothoid curves for controlling the shape of variable-length continuous manipulators.
Shape Servoing of a Soft Object Using Fourier Series and a Physics-Based Model
Fouad Makiyeh, Alexandre Krupa
OptimizationRobotic IntelligenceImagePhysics Related
🎯 What it does: Proposed a physics-based robot controller that deforms soft objects into desired three-dimensional shapes using a finite number of control points, and represents the shapes with Fourier descriptors.
Shared Autonomy Control for Slosh-Free Teleoperation
Rafael I. Cabral Muchacho, Sami Haddadin
OptimizationRobotic IntelligenceAgentic AI
🎯 What it does: Proposed a real-time, non-grasping splash-free teleoperation control framework and motion generator to achieve splash-free teleoperation of liquids.