IROS 2025 Papers — Page 11
IEEE/RSJ International Conference on Intelligent Robots and Systems · 1984 papers
Landmark-Based Goal Recognition for Shared Autonomy: A Framework for Enhanced Teleoperation
Guillaume Lorthioir, I. Ramirez-Alpizar
RecognitionRobotic IntelligenceRecurrent Neural Network
🎯 What it does: Proposes a framework applicable to various robots, particularly humanoid robots, for shared autonomy. The framework achieves target recognition through Bayesian filtering and Hidden Markov Models (HMMs), and employs a landmark heuristic to enable low-computation observation likelihood estimation. Assistance is dynamically provided based on confidence levels.
LaneMind: Seeing Lanes Like Human Drivers
Zhengyan Qian, Qian Ma
Autonomous DrivingTransformer
🎯 What it does: Propose the LaneMind framework to achieve lane detection by integrating human visual perception with geometric modeling.
Lanes Are Not Enough: Enhancing Trajectory Prediction in Intralogistics Through Detailed Environmental Context
Alexander Prutsch, Horst Possegger
Autonomous Driving
🎯 What it does: Propose the use of a new environmental context encoder module in internal logistics to improve trajectory prediction, and can be seamlessly integrated into advanced autonomous driving systems.
LangGrasp: Leveraging Fine-Tuned LLMs for Language Interactive Robot Grasping with Ambiguous Instructions
Yunhan Lin, Huasong Min
Robotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningVision-Language-Action ModelImagePoint Cloud
🎯 What it does: Propose the LangGrasp framework, which utilizes fine-tuned LLM and point cloud localization to achieve language-interactive grasping
Language as Cost: Proactive Hazard Mapping using VLM for Robot Navigation
Mintaek Oh, Seong-Woo Kim
Robotic IntelligenceLarge Language ModelVision Language ModelMultimodality
🎯 What it does: Proposes a zero-shot language cost mapping framework based on VLM for predicting dynamic hazards and assigning risk costs to robot navigation, achieving active obstacle avoidance by integrating with geometric obstacle maps.
Language-Guided Hierarchical Planning with Scene Graphs for Tabletop Object Rearrangement
Wooseok Oh, Songhwai Oh
Robotic IntelligenceGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: Propose a hierarchical planning framework based on scene graphs for robot desktop object rearrangement, which generates scene graphs using language goals and plans transitions from the current scene graph to the target scene graph, ultimately generating executable low-level plans to achieve configurations that meet language goals and are well-arranged.
Large Language Model-Based Robot Task Planning from Voice Command Transcriptions
Afonso Certo, Pedro U. Lima
Robotic IntelligenceLarge Language ModelText
🎯 What it does: Proposed a complete pipeline that leverages large language models (LLMs) to directly convert spoken instructions into coherent action plans, integrating environmental context into model inputs to generate more efficient and context-aware plans.
Large-Scale Mixed-Traffic and Intersection Control using Multi-agent Reinforcement Learning
Songyan Liu, Shuai Li
Autonomous DrivingOptimizationReinforcement LearningGraph
🎯 What it does: Control of large-scale mixed traffic networks using decentralized multi-agent reinforcement learning, where some intersections are managed by traffic signals and others are controlled by robot vehicles.
LATMOS: Latent Automaton Task Model from Observation Sequences
Weixiao Zhan, N. Atanasov
Auto EncoderWorld ModelSequential
🎯 What it does: Proposes a potential automaton task model named LATMOS based on observation sequences, which automatically decomposes tasks, understands the current task state, and supports planning and verification from observations of correctly executed tasks.
Layer Decomposition and Morphological Reconstruction for Task-Oriented Infrared Image Enhancement
Siyuan Chai, Tong Liu
RestorationImage
🎯 What it does: Propose a task-oriented infrared image enhancement method, including salient information extraction through l0-l1 layer decomposition and morphological reconstruction.
LBAP: Improved Uncertainty Alignment of LLM Planners using Bayesian Inference
James F. Mullen, Dinesh Manocha
Robotic IntelligenceTransformerLarge Language Model
🎯 What it does: Propose the LBAP method, which aligns uncertainty in robot planning by leveraging off-the-shelf large language models (LLMs) combined with Bayesian inference, to reduce hallucinations and minimize human intervention.
LBE-DDIK: Is One Model Good Enough to Learn-By-Example the Inverse Kinematics of Multiple Serial Robots?
Jacket Demby's, G. DeSouza
Robotic IntelligenceTransformer
🎯 What it does: Proposed and evaluated a single neural network model for learning inverse kinematics (IK) in multi-robot systems (6 and 7 degrees of freedom (DoF)) using the Learning-By-Example (LBE) framework.
LDexMM: Language-Guided Dexterous Multi-Task Manipulation with Reinforcement Learning
Hengxu Yan, Cewu Lu
Robotic IntelligenceReinforcement LearningVision-Language-Action ModelImageText
🎯 What it does: Proposes a language-guided multi-task flexible grasping and manipulation framework (LDexMM), which uses language instructions to guide a segmentation model in generating grasp poses for functional parts of objects, then employs reinforcement learning to further refine the grasp posture and complete the task, while using language constraints to focus on specified objects.
Learning Accurate Whole-body Throwing with High-frequency Residual Policy and Pullback Tube Acceleration
Yuntao Ma, Marco Hutter
OptimizationRobotic Intelligence
🎯 What it does: This paper proposes a whole-body throwing control framework combining learning and model-based control for legged mobile robots to perform grasping and throwing, and implements the framework on hardware.
Learning Adaptive Dexterous Grasping from Single Demonstrations
Liangzhi Shi, Hao Su
Robotic IntelligenceReinforcement LearningVision Language ModelImageText
🎯 What it does: Propose the AdaDexGrasp framework, which learns a grasping skill library from a single human demonstration and selects appropriate skills via a Vision-Language Model (VLM) based on user instructions. It combines trajectory following reward and curriculum learning to improve RL sample efficiency, and finally validates its effectiveness in simulation and real environments, achieving zero-shot transfer to the PSYONIC Ability Hand.
Learning Appearance and Motion Cues for Panoptic Tracking
Juana Valeria Hurtado, Abhinav Valada
Object TrackingConvolutional Neural Network
🎯 What it does: Propose a panoptic tracking method that simultaneously captures semantic information, instance-level appearance, and motion features
Learning Dexterous In-Hand Manipulation with Multifingered Hands via Visuomotor Diffusion
Piotr Koczy, Danica Kragic
Robotic IntelligenceDiffusion model
🎯 What it does: Proposes a framework for learning multi-finger wrist inner-side operations using audio-visual diffusion strategies, capable of performing complex in-hand operations such as unscrewing a bottle cap with one hand;
Learning Distributed End-to-End Hunting Locomotion for Multiple Quadruped Robots
Chung Yui Yeung, Chin Pang Ho
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a distributed end-to-end reinforcement learning framework that enables multiple quadrupedal robots to learn motion control for collaborative hunting behavior.
Learning Flow-Adaptive Dynamic Model for Robotic Fish Swimming in Unknown Background Flow
Kaitian Chao, Yang Wang
Domain AdaptationRobotic Intelligence
🎯 What it does: Proposes a data-driven dynamic modeling framework for characterizing the swimming motion of biomimetic fish in unknown background flows without requiring explicit flow information.
Learning Force Distribution Estimation for the GelSight Mini Optical Tactile Sensor Based on Finite Element Analysis
E. Helmut, Jan Peters
Robotic IntelligenceConvolutional Neural NetworkImagePhysics Related
🎯 What it does: This paper proposes using a U-net network to directly predict normal and shear force distributions from raw images of the GelSight Mini optical tactile sensor.
Learning from Human Conversations: A Seq2Seq based Multi-modal Robot Facial Expression Reaction Framework in HRI
Zhegong Shangguan, Adriana Tapus
Robotic IntelligenceRecurrent Neural NetworkMultimodality
🎯 What it does: Using a Seq2Seq framework and a motion mapping model driven by deep neural networks, human speech and facial expressions are mapped to facial actions executable by robots, enabling robots to generate self-reactive facial expressions.
Learning Generalizable 3D Manipulation With 10 Demonstrations
Yu Ren, Huijie Fan
Robotic IntelligenceDiffusion modelImage
🎯 What it does: Learn 3D manipulation strategies using only 10 demonstrations and achieve robust generalization to unseen spatial configurations.
Learning Generalizable Feature Fields for Mobile Manipulation
Ri-Zhao Qiu, Xiaolong Wang
Data SynthesisRobotic IntelligenceNeural Radiance FieldContrastive Learning
🎯 What it does: Proposed a generic feature field GeFF that enables real-time unified navigation and manipulation representation on mobile robots.
Learning Generalizable Language-Conditioned Cloth Manipulation from Long Demonstrations
Han Zhao, Xueqian Wang
Robotic IntelligenceTransformerLarge Language ModelAgentic AI
🎯 What it does: Proposed a pipeline for autonomously learning basic skills from long demonstrations and combining them to accomplish unseen tasks;
Learning Gentle Grasping Using Vision, Sound, and Touch
Ken Nakahara, Roberto Calandra
Robotic IntelligenceVision-Language-Action ModelImageMultimodalityAudio
🎯 What it does: By combining visual, tactile, and auditory signals, an action-conditioned model is trained to predict the stability and gentleness of future grasp candidates from raw visual-tactile inputs, and select the optimal grasp action based on these predictions.
Learning Goal-Directed Object Pushing in Cluttered Scenes With Location-Based Attention
Nils Dengler, Maren Bennewitz
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a method for goal-oriented object pushing in cluttered scenes using position attention.
Learning Manipulation Skills through Robot Chain-of-Thought with Sparse Failure Guidance
Kaifeng Zhang, Yang Gao
Robotic IntelligenceReinforcement LearningVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Proposes an algorithm that decomposes tasks into finer-grained subtasks and uses a vision-language model (VLM) to provide reward guidance, while introducing a VLM self-imitation learning process to accelerate learning.
Learning Natural and Robust Hexapod Locomotion over Complex Terrains via Motion Priors based on Deep Reinforcement Learning
Xin Liu, Feng Gao
Robotic IntelligenceReinforcement LearningGenerative Adversarial Network
🎯 What it does: Proposed a deep reinforcement learning method based on motion priors, achieving natural and robust gait learning for a real six-legged robot;
Learning Object Compliance via Young’s Modulus from Single Grasps using Camera-Based Tactile Sensors
Michael Burgess, Laurence Willemet
Robotic IntelligenceImage
🎯 What it does: Estimate the object's elastic modulus (Young's modulus) to characterize compliance through a single parallel grasp using a camera-based tactile sensor, joint pose, and force measurement.
Learning Perceptive Humanoid Locomotion over Challenging Terrain
Wandong Sun, Hong Liu
Knowledge DistillationRobotic IntelligenceWorld Model
🎯 What it does: Train a world model capable of denoising sensor noise and estimating states through a teacher-student distillation framework, to enhance perception and gait planning for humanoid robots on uneven terrain.
Learning Point Correspondences In Radar 3D Point Clouds For Radar-Inertial Odometry
J. Michalczyk, J. Steinbrener
Pose EstimationTransformerSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Designed a Transformer-based learning framework for predicting robust point correspondences in noise-sparse SoC FMCW radar 3D point clouds.
Learning Predictive Control with Online Modeling for Agile Maneuvering of Autonomous Vehicles
Xin Yin, Xinglong Zhang
Autonomous DrivingOptimizationReinforcement LearningWorld Model
🎯 What it does: Propose a closed-loop MPC strategy combining learning-based adaptive model predictive control (AMPC) with Actor-Critic learning (ACL), using neural networks to online model vehicle dynamic uncertainties and synchronously update control policies and models to achieve agile maneuvering control.
Learning Robust Agile Flight Control with Stability Guarantees
Lukas Pries, M. Ryll
Robotic Intelligence
🎯 What it does: Proposed and verified a neural-enhanced feedback controller for achieving precise trajectory tracking of high-speed agile quadrotors under platform limit states.
Learning Robust and Flexible Locomotion of Wheel-Legged Quadruped Robots in Complex Terrains
Shiyu Zhou, Rongrong Wang
Robotic IntelligenceReinforcement LearningTime Series
🎯 What it does: Proposed a reinforcement learning-based control framework for robust and flexible locomotion of single-unit wheeled-legged quadruped robots on complex terrains.
Learning Smooth Humanoid Locomotion through Lipschitz-Constrained Policies
Zixuan Chen, Xue Bin Peng
OptimizationRobotic Intelligence
🎯 What it does: Proposed and verified a method to achieve smooth humanoid robot gait control by imposing Lipschitz constraints on the learning policy, termed Lipschitz-Constrained Policies (LCP)
Learning Symmetric Legged Locomotion via State Distribution Symmetrization
Chengrui Zhu, Yong Liu
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a learning framework that systematically optimizes the gait symmetry of multi-legged and humanoid robots through state distribution symmetrization
Learning Through Retrospection: Improving Trajectory Prediction for Automated Driving with Error Feedback
Steffen Hagedorn, Alexandru Condurache
Autonomous DrivingVideoPoint CloudSequential
🎯 What it does: Proposed a backtracking technique that utilizes closed-loop replay training to enable the model to reflect on previous predictions and correct errors during the prediction process, thereby improving trajectory prediction quality.
Learning to Exploit Leg Odometry Enables Terrain-Aware Quadrupedal Locomotion
Yong Zhou, Zengmao Wang
Robotic IntelligenceReinforcement LearningSimultaneous Localization and Mapping
🎯 What it does: A lightweight framework utilizing only the quadruped robot's own sensors, combining learning-based leg odometry and reinforcement learning-trained locomotion strategies, achieving terrain perception and utilization;
Learning to Generate Vectorized Maps at Intersections with Multiple Roadside Cameras
Quanxin Zheng, Haoyi Xiong
GenerationAutonomous DrivingImage
🎯 What it does: Propose an end-to-end visual network called MRC-VMap based on multi-roadside cameras, designed to directly generate high-precision vectorized maps at intersections.
Learning to Hang Crumpled Garments with Confidence-Guided Grasping and Active Perception
Shengzeng Huo, D. Navarro-Alarcón
Depth EstimationRobotic Intelligence
🎯 What it does: Proposed a confidence-guided grasping strategy and a two-step suspension method, utilizing dual-arm active perception to achieve suspension of wrinkled garments.
Learning to Initialize Trajectory Optimization for Vision-Based Autonomous Flight in Unknown Environments
Yicheng Chen, Qingdong Li
OptimizationRobotic IntelligenceImage
🎯 What it does: Proposes a neural network enhanced trajectory planner (NEO-Planner), which learns to predict spatial and temporal parameters from raw visual sensor observations, providing heuristic initial values for non-convex trajectory optimization to accelerate planning and support robust online replanning.
Learning to Perform Low-Contact Autonomous Nasotracheal Intubation by Recurrent Action-Confidence Chunking with Transformer
Yu Tian, Hongliang Ren
Safty and PrivacyRobotic IntelligenceRecurrent Neural NetworkTransformerImageBiomedical Data
🎯 What it does: Developed an autonomous nasal intubation system, including a prosthesis with embedded force sensors and the RACCT model for handling pipe-tissue interactions.
Learning to Solve the Multi-Agent Task Assignment Problem for Automated Data Centers
Christelle Loiodice, J. Andreoli
OptimizationReinforcement Learning
🎯 What it does: The study addresses the multi-agent task allocation problem of mobile robots collaborating with humans to complete installation and maintenance tasks in large data centers, formalizing it as a Markov Decision Process (MDP) and proposing an end-to-end learning method to solve it.
Learning to Throw-Flip
Yang Liu, A. Billard
Robotic Intelligence
🎯 What it does: Propose a method for a robot to accurately throw and flip objects to a target landing posture.
Learning to traverse challenging terrain using vision and forward kinematics*
Jiajun Dong, Feng Dong
Robotic IntelligenceSupervised Fine-TuningImage
🎯 What it does: Propose a visual locomotion controller for quadruped robots that combines vision with forward kinematics to enhance their walking ability on challenging terrains
Learning Upright and Forward-Facing Object Poses using Category-level Canonical Representations
Bing Han, Xuefeng Chen
Pose EstimationRepresentation LearningContrastive Learning
🎯 What it does: Proposes a category-level self-supervised canonicalization method to construct a unified upright and forward-facing 3D object pose representation.
Learning Whole-Body Control for Small-Sized Quadruped Robots with a Flexible Spine
Dixuan Jiang, Qing Shi
OptimizationRobotic Intelligence
🎯 What it does: Proposes a teacher-student online learning framework for whole-body control of small quadruped robots (with a flexible spine).
Learning-based Keypoints Detection with Topological Order on Deformable Linear Objects from Incomplete Point Clouds
Can Li, Lei Sun
Pose EstimationRecurrent Neural NetworkPoint Cloud
🎯 What it does: Propose a deep learning-based method that utilizes the topological properties of DLO to detect key points in incomplete point clouds while maintaining topological order.
Learning-Based Motion Controller for Reconfigurable Microswarms
Yamei Li, Lidong Yang
Robotic IntelligenceTime Series
🎯 What it does: Proposed and implemented a motion control framework for magnetic microrobot swarms based on Learning from Demonstration (LfD).
Learning-Based Passive Fault-Tolerant Control of a Quadrotor with Rotor Failure
Jiehao Chen, Yunjiang Lou
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a learning-based passive fault-tolerant control method that can handle any single rotor failure without requiring rotor fault information or controller switching.
Learning-Based Predictive Impedance Control Towards Safe Predefined-Time Physical Robotic Interaction
Junyuan Xue, Tong Heng Lee
OptimizationSafty and PrivacyRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposed an MPC-based impedance control framework that combines learning regulation to achieve predefined time convergence, and dynamically selects task-oriented and safety-oriented impedance models at the lower level to enhance performance and safety.
Learning-Based Quadruped Robot Framework for Locomotion on Dynamic Rigid Platforms
Kai-Hsiang Huang, Tianjiang Hu
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a reinforcement learning-based framework for controlling quadruped robot locomotion on dynamic rigid platforms (e.g., ship vibrations);
Least Commitment Planning for the Object Scouting Problem
Max Merlin, D. Paulius
OptimizationRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposed a specialized reconnaissance partial-order planner (SPOP) for LOMDP, utilizing partial-order and regression planning to address knowledge gaps in robots' understanding of object presence, location, and states in the environment.
Legged Robot State Estimation Using Invariant Neural-Augmented Kalman Filter with a Neural Compensator
Seokju Lee, Kyung-Soo Kim
Robotic Intelligence
🎯 What it does: Proposed a neural network augmented invariant Kalman filter designed on the Lie group structure to improve state estimation for legged robots.
LEGO-Motion: Learning-Enhanced Grids with Occupancy Instance Modeling for Class-Agnostic Motion Prediction
Kangan Qian, Diange Yang
Autonomous DrivingTransformerPoint CloudBenchmark
🎯 What it does: Proposes the LEGO-Motion framework, combining instance-level reasoning with occupancy grid modeling to achieve class-agnostic motion prediction.
LensDFF: Language-enhanced Sparse Feature Distillation for Efficient Few-Shot Dexterous Manipulation
Qian Feng, Alois Knoll
Knowledge DistillationRobotic IntelligenceVision-Language-Action ModelImagePoint Cloud
🎯 What it does: Proposes the language-enhanced sparse feature distillation framework LensDFF for efficient single-view few-shot generalization of sparse feature fields, and constructs a few-shot dexterous manipulation framework based on this framework
LERa: Replanning with Visual Feedback in Instruction Following
S. Pchelintsev, Aleksandr I. Panov
Robotic IntelligenceVision Language ModelImageTextMultimodality
🎯 What it does: Proposes LERa, a vision-language model-based re-planning method that utilizes visual feedback to re-plan after task execution failure. The process includes observation (generating scene descriptions and identifying errors), explanation (providing corrective guidance), and re-planning (modifying the plan)
Let Me Show You: Learning by Retrieving from Egocentric Video for Robotic Manipulation
Yichen Zhu, Feifei Feng
Robotic IntelligenceVision-Language-Action ModelVideoRetrieval-Augmented Generation
🎯 What it does: Propose a method that learns robot manipulation strategies by retrieving task-related videos from external sources and fusing intermediate layer information.
Leveraging Semantic Graphs for Efficient and Robust LiDAR SLAM
Neng Wang, Xieyuanli Chen
Autonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed an efficient and robust LiDAR SLAM framework called SG-SLAM based on semantic maps.
Leveraging Temporally Extended Behavior Sharing for Multi-task Reinforcement Learning
Gawon Lee, H. Kim
Reinforcement Learning
🎯 What it does: Propose MT-Lévy, a novel exploration strategy that improves sample efficiency in multi-task reinforcement learning environments by combining cross-task behavior sharing with Lévy flight-inspired temporal extension exploration.
Leveraging Text-Driven Semantic Variation for Robust OOD Segmentation
Seungheon Song, Jaekoo Lee
SegmentationAnomaly DetectionTransformerPrompt EngineeringVision Language ModelImageMultimodality
🎯 What it does: This paper proposes and trains a text-driven OOD segmentation model, combining visual-language encoders, Transformer decoders, semantic distance-based OOD prompts, and OOD semantic augmentation to learn diverse objects in the visual-language space and achieve robust segmentation for unknown objects.
LGDD: Local-Global Synergistic Dual-Branch 3D Object Detection Using 4D Radar
Xiaokai Bai, Hui-Liang Shen
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: Proposed a local-global collaborative dual-branch 3D object detection framework based on 4D radar, named LGDD, which includes a point-based branch (employing a voxel attention point feature extractor and cluster voting) and a pillar-based branch (utilizing query-based feature pre-fusion and proposal masks), and achieves alignment between local instances and global context through a semantic-geometry awareness fusion module.
LGNav: Zero-Shot Object Navigation Driven by Language and Pointing Gesture Using Large Vision-Language Models
Weiyi Zhu, Zhehan Yang
Object DetectionDepth EstimationAutonomous DrivingLarge Language ModelVision Language ModelVision-Language-Action ModelTextMultimodality
🎯 What it does: Proposes the LGNav framework to achieve zero-shot language and pointing gesture-driven object navigation.
LGPR: Local Feature Learning Brings More Generalizable Visual Place Recognition
Shuai Su, Qi Chen
RetrievalRepresentation LearningConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposes a visual place recognition (VPR) framework with a shared lightweight keypoint extraction module, which jointly learns local keypoint extraction and VPR by utilizing self-attention and cross-attention mechanisms to fuse irregularly distributed key features.
LHMM: A Tightly-Coupled LiDAR-Inertial Hybrid-Map Matching Approach for Robust and Efficient Global Localization
Junyuan Lu, Yu Zhang
Autonomous DrivingSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Propose a tightly coupled LiDAR-IMU hybrid mapping matching framework called LHMM, which first compresses the prior map through skeletonization, and then jointly optimizes IMU, skeleton feature prior maps, and local voxel maps under a MAP estimation framework to achieve single-step pose estimation and prevent error propagation; the local map adopts a hole-aware keyframe mechanism, focusing on areas with environmental changes or under-mapped regions to reduce computational load.
LI-SLAM: Lightweight and Incremental Semantic Visual Localization and Mapping for Autonomous Valet Parking
Huateng Wu, Song Zhao
Autonomous DrivingSimultaneous Localization and MappingImageMultimodality
🎯 What it does: Proposes a method for directly regressing semantic corners to build a semantic map, and introduces a map update and merging strategy insensitive to environmental and temporal changes. Utilizes four-camera synthesized panoramic images, IMU, and wheel encoders to construct a global semantic map, and verifies localization accuracy through real-world experiments.
LiDAR-IMU Fusion System with Adaptive Scanning for High-Resolution Deformation Monitoring of Underground Infrastructures
Menggang Li, Gongbo Zhou
SegmentationOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Developed a high-resolution deformation monitoring system for underground coal mine surrounding rock based on LiDAR-IMU fusion and adaptive scanning.
LiDAR-Inertial Odometry in Dynamic Driving Scenarios using Label Consistency Detection
Zikang Yuan, Xin Yang
Autonomous DrivingComputational EfficiencySimultaneous Localization and MappingPoint Cloud
🎯 What it does: Propose a LiDAR-inertial odometry method for dynamic driving scenarios that eliminates the impact of moving objects by constructing binary labels and utilizing label differences to detect moving objects.
Lifelong Morphology Learning for Deformable Embodied Agents
Yinsong Wang, Huaping Liu
OptimizationRobotic IntelligenceMeta LearningBenchmark
🎯 What it does: Proposed the Ske-Ex framework, achieving lifelong morphological learning for deformable agents that can continuously adapt to diverse terrains.
LightPlanner: Unleashing the Reasoning Capabilities of Lightweight Large Language Models in Task Planning
Weijie Zhou, Jinqiao Wang
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose LightPlanner, which improves complex task planning by leveraging the inference capabilities of lightweight LLMs, employing parameterized function calls, hierarchical deep reasoning, and memory modules to achieve more efficient and precise action control.
Lightweight Temporal Transformer Decomposition for Federated Autonomous Driving
Tuong Khanh Long Do, Anh Nguyen
Autonomous DrivingFederated LearningTransformerImageTime Series
🎯 What it does: Propose a lightweight time Transformer decomposition method, which splits large attention maps into smaller matrices to process continuous image frames and steering sequences, reduces model complexity, and achieves real-time autonomous driving prediction in federated learning environments.
LiHRA: A LiDAR-Based HRI Dataset for Automated Risk Monitoring Methods
Frederik Plahl, Andrey Morozov
Safty and PrivacyRobotic IntelligenceMultimodalityPoint CloudBenchmark
🎯 What it does: Proposed the LiHRA dataset and a context-based risk monitoring method
Like Playing a Video Game: Spatial-Temporal Optimization of Foot Trajectories for Controlled Football Kicking in Bipedal Robots
Wanyue Li, Peng Lu
OptimizationRobotic Intelligence
🎯 What it does: Designed a space-time optimized foot trajectory for bipedal robot soccer kicking
LIM: A Low-Complexity Local Feature Image Matching Network for Real-Time Embedded Applications
Shanquan Ying, Junjie Dai
RetrievalComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: Proposed a lightweight image matching network that achieves a balance between accuracy and efficiency, and improves robustness to large image rotations.
LiMo-Calib: On-Site Fast LiDAR-Motor Calibration for Quadruped Robot-Based Panoramic 3D Sensing System
Jianping Li, Lihua Xie
OptimizationRobotic IntelligenceSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed and validated a field rapid LiDAR-servo calibration method LiMo-Calib without external targets, applicable to panoramic 3D perception systems on quadruped robot platforms.
Lip Geometry-Constrained Smooth Sliding Path Planning for Robotic Negative Pressure Therapy on Extremities
Zihao Li, Xin-Jun Liu
OptimizationRobotic Intelligence
🎯 What it does: A smooth sliding path planning method for robotic negative pressure continuous suction therapy on limb edema was proposed. By comparing point normal vectors with the suction head angle to identify easily sealed areas, the path planning was simplified to a two-dimensional plane. The conjugate gradient method was used to optimize the path under constraints of centroid distance and smoothness, ultimately generating a smooth sliding path with suction head pressure instructions. Ten sliding suction experiments were conducted on a robotic hand with variable-sized suction cups, successfully achieving six sliding suction operations on a simulated arm.
LITE: A Learning-Integrated Topological Explorer for Multi-Floor Indoor Environments
Junhao Chen, Yong Liu
Object DetectionRobotic IntelligenceConvolutional Neural NetworkSimultaneous Localization and MappingImage
🎯 What it does: Propose a learning-integrated topology explorer called LITE, specifically designed for exploration in multi-level indoor environments.
LiteFat: Lightweight Spatio-Temporal Graph Learning for Real-Time Driver Fatigue Detection
Jing Ren, Feng Xia
ClassificationAutonomous DrivingGraph Neural NetworkVideo
🎯 What it does: Proposed a lightweight spatiotemporal graph learning model called LiteFat for real-time driver fatigue detection; video streams are converted into spatiotemporal graphs through facial keypoint detection, facial features are extracted using MobileNet to construct a feature matrix, and then a lightweight spatiotemporal graph neural network is employed to identify fatigue signs.
LITHE-joint: Variable Stiffness Compliant Spherical Contact Joint in an Under-Actuated System
Sanpoom Punapanont, P. Manoonpong
Robotic Intelligence
🎯 What it does: Proposed and implemented a variable stiffness spherical contact joint called LITHE-joint for underactuated systems.
LLA-MPC: Fast Adaptive Control for Autonomous Racing
Maitham F. Al-Sunni, John M. Dolan
Autonomous Driving
🎯 What it does: Proposes the LLA-MPC framework, achieving fast adaptive control to handle rapid changes in tire-road interaction, applicable to autonomous racing.
LLM-CBT: LLM-Driven Closed-Loop Behavior Tree Planning for Heterogeneous UAV-UGV Swarm Collaboration
Yuanyuan Tian, Yabo Liu
Robotic IntelligenceTransformerLarge Language Model
🎯 What it does: Proposed a closed-loop behavior tree planning framework based on LLM, named LLM-CBT, for heterogeneous UAV-UGV unmanned swarm task planning.
LLM-Driven Hierarchical Planning: Long-horizon Task Allocation for Multi-Robot Systems in Cross-Regional Environments
Yachao Wang, M. Q. Meng
OptimizationRobotic IntelligenceTransformerLarge Language ModelText
🎯 What it does: Proposes a three-phase task planning framework for multi-robot cross-regional long-term composite task planning, addressing spatial semantic understanding of natural language tasks and subtask allocation optimization.
LLM-Informed Iterative Planning for Object Search and Relocation in Indoor Environments
Taxiarchis-Foivos Blounas, G. Nikolakopoulos
Robotic IntelligenceTransformerLarge Language Model
🎯 What it does: Propose an indoor object search and relocation method based on large language model (LLM) iterative planning
LLplace: Embodied 3D Indoor Layout Synthesis Framework with Large Language Model
Yixuan Yang, Feng Zheng
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose LLplace, based on the lightweight open-source LLM Llama3, using user input room types and required items through conversational interaction to generate and dynamically edit 3D interior layouts.
LMMCoDrive: Cooperative Driving with Large Multimodal Models
Haichao Liu, Jun Ma
Autonomous DrivingOptimizationTransformerLarge Language ModelImageMultimodality
🎯 What it does: Propose the LMMCoDrive framework, which integrates scheduling and motion planning by leveraging Large Multimodal Models (LMM) combined with bird's-eye view (BEV) representations, and enhances traffic efficiency and passenger experience in AMoD systems through decentralized safety-constrained optimization via ADMM.
Local Path Optimization in The Latent Space Using Learned Distance Gradient
Jiawei Zhang, Jifeng Guo
OptimizationRobotic Intelligence
🎯 What it does: Train neural networks to predict the minimum distance between robots and obstacles, and use the learned distance gradient in the latent space to guide robot motion, thus proposing a local path optimization algorithm that integrates path validity checks to reduce replanning time.
Localization of an Unmanned Underwater Vehicle Using a Tethered Cooperative Surface Vehicle and Hybrid EKF/Grid-Based Method
A. M. Oxford, Brendan Englot
Robotic IntelligenceSimultaneous Localization and MappingMultimodality
🎯 What it does: Collaborative use of a cable-equipped unmanned surface vehicle and an unmanned underwater vehicle (UUV) for UUV localization, leveraging information from cameras, sonar, and cable data.
LOG-SLAM: Large-Scale Outdoor Gaussian SLAM for Dense Mapping and Loop Closure in Kilometer-Scale Scene Reconstruction
Long Wang, Dan Luo
Pose EstimationDepth EstimationAutonomous DrivingGaussian SplattingSimultaneous Localization and MappingVideoPoint Cloud
🎯 What it does: Propose LOG-SLAM, utilizing Gaussian Splatting for tracking and map building in large outdoor environments
LoL-NMPC: Low-Level Dynamics Integration in Nonlinear Model Predictive Control for Unmanned Aerial Vehicles
Parakh M. Gupta, M. Saska
OptimizationPhysics Related
🎯 What it does: A new nonlinear model predictive controller, LoL-NMPC, is studied, which explicitly incorporates low-level flight controllers (e.g., PID) and motor dynamics into the model to enhance the accuracy and robustness of drones in high-speed agile trajectory tracking.
LOMORO: Long-term Monitoring of Dynamic Targets with Minimum Robotic Fleet under Resource Constraints
Mingke Lu, Meng Guo
Robotic Intelligence
🎯 What it does: Proposed an online collaborative monitoring scheme called LOMORO for achieving long-term monitoring, path routing, and resource recharging of dynamic targets under resource-constrained conditions;
Long-Distance Delivery of Collective Cell Microrobots Driven by Mobile Magnetic Actuation System
Yimin Sun, Qianqian Wang
Robotic Intelligence
🎯 What it does: Investigated the long-distance delivery of collective cell microrobots under flowing conditions using a mobile magnetic field driving system, proposed a magnetic driving strategy, and conducted experimental validation.
Long-horizon Locomotion and Manipulation on a Quadrupedal Robot with Large Language Models
Yutao Ouyang, Yi Wu
Robotic IntelligenceTransformerLarge Language ModelReinforcement Learning
🎯 What it does: Proposed and implemented a system based on large language models (LLM) that enables quadruped robots to achieve end-to-end solutions for long-term tasks through high-level semantic planning and low-level reinforcement learning, generating hybrid discrete-continuous plans and converting them into executable robot code;
Look Before You Leap: Using Serialized State Machine for Language Conditioned Robotic Manipulation
Tong Mu, Mehran Armand
Robotic IntelligenceVision-Language-Action Model
🎯 What it does: Propose a framework that utilizes serialized finite state machines (FSM) to generate demonstration data, aiming to improve the success rate of language-conditioned robot manipulation tasks, particularly those requiring precise long-sequence interactions.
LoopSR: Looping Sim-and-Real for Lifelong Policy Adaptation of Legged Robots
Peilin Wu, Weinan Zhang
Domain AdaptationRobotic IntelligenceTransformerReinforcement LearningAuto EncoderContrastive LearningSequential
🎯 What it does: Propose the LoopSR framework, which uses a transformer encoder to map real-world trajectories to a latent space, reconstruct a digital twin, and continuously improve reinforcement learning strategies for legged robots after deployment.
Low-effort Iterative Dataset Generation Pipeline for Unknown Object Instance Segmentation
Florian Jordan, Marco F. Huber
SegmentationData SynthesisTransformerSupervised Fine-TuningPrompt EngineeringImage
🎯 What it does: Proposes a low-cost, low-human-intervention RGB-only pipeline to iteratively place unknown objects in scenes for generating instance segmentation datasets and evaluates multiple change detection-based algorithms.
Low-Fidelity Visuo-Tactile Pre-Training Improves Vision-Only Manipulation Performance
Selam Gano, A. Farimani
Representation LearningRobotic IntelligenceReinforcement LearningImage
🎯 What it does: The study uses low-cost BeadSight tactile sensors for pre-training and employs only visual input in downstream tasks to enhance operational performance.
Low-Latency Privacy-Aware Robot Behavior guided by Automatically Generated Text Datasets
Yuta Irisawa, Ken Sakurada
Data SynthesisSafty and PrivacyRobotic IntelligenceLarge Language ModelVision Language ModelImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: Automatically construct a privacy text dataset and utilize a multimodal retrieval method to identify whether images infringe on privacy without using sensitive images for training, thereby achieving low-latency robot privacy-aware behavior
LPVIMO-SAM: Tightly-coupled LiDAR/Polarization Vision/Inertial/Magnetometer/Optical Flow Odometry via Smoothing and Mapping
Derui Shan, Du Tao
Autonomous DrivingOptimizationSimultaneous Localization and MappingOptical FlowImageMultimodalityPoint Cloud
🎯 What it does: Proposed and implemented a tightly coupled LPVIMO-SAM framework that integrates LiDAR, polarized vision, inertial measurement unit, magnetometer, and optical flow to achieve high-precision real-time state estimation and map building.
LR2Depth: Large-Region Aggregation at Low Resolution for Efficient Monocular Depth Estimation
Chao Ning, Naoto Yokoya
Depth EstimationComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: Proposes a monocular depth estimation method that uses large convolutional kernels on low-resolution feature maps for large-scale feature aggregation
LS-HAR: Language Supervised Human Action Recognition with Salient Fusion, Construction Sites as a Use-Case
Mohammad Mahdavian, Mo Chen
RecognitionTransformerLarge Language ModelPrompt EngineeringVision-Language-Action ModelVideoTextMultimodality
🎯 What it does: The LS-HAR method, which uses language supervision to combine skeleton and visual information for human action recognition.