arXivSub Start free trial

ICRA 2025 Papers — Page 10

IEEE International Conference on Robotics and Automation · 1604 papers

Masked Sensory-Temporal Attention for Sensor Generalization in Quadruped Locomotion

Dikai Liu, Simon See

Robotic IntelligenceTransformerTime Series

🎯 What it does: Proposes the Masked Sensory-Temporal Attention (MSTA) mechanism, which leverages Transformer-based masked attention to directly focus on the sensor level, thereby enhancing perception-temporal understanding and the ability to process different sensor combinations in quadruped robot locomotion.

Master Rules from Chaos: Learning to Reason, Plan, and Interact from Chaos for Tangram Assembly

Chao Zhao, Qifeng Chen

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Studies the robot puzzle assembly problem and proposes the MRChaos learning framework, which can self-explore and learn assembly strategies.

Mastering Agile Jumping Skills from Simple Practices with Iterative Learning Control

Chuong Nguyen, Quan Nguyen

Robotic Intelligence

🎯 What it does: Improve simulation accuracy and safety by gradually learning from simple jumping tasks through Iterative Learning Control (ILC), and realize various more challenging jumping tasks on hardware.

Match Policy: A Simple Pipeline from Point Cloud Registration to Manipulation Policies

Hao-zhe Huang, Robert Platt

Robotic IntelligencePoint CloudBenchmark

🎯 What it does: Proposed a simple pipeline from point cloud registration to manipulation strategies for solving high-precision grasping and placing tasks.

MatchMaker: Automated Asset Generation for Robotic Assembly

Yian Wang, Iretiayo Akinola

GenerationData SynthesisRobotic Intelligence

🎯 What it does: Proposed a pipeline for automatically generating diverse, simulation-ready assembly asset pairs.

Materials Matter: Investigating Functional Advantages of Bio-Inspired Materials via Simulated Robotic Hopping

Andrew K. Schulz, Maegan Tucker

Robotic IntelligencePhysics Related

🎯 What it does: Investigate the impact of non-rigid materials on the functional performance of bouncing robots, and demonstrate the advantages of material gradient legs through simulation

Mathematical Modeling and Rolling Motion Generation of Planar Seven-link Robot That Forms Passive Closed and Active Open Chains

Fumihiko Asano, Isao T. Tokuda

Robotic IntelligencePhysics RelatedOrdinary Differential Equation

🎯 What it does: Studied the mathematical modeling and basic motion characteristics of a planar seven-bar robot under two structural configurations: passive closed chain and active open chain, and verified its motion behavior under different slope angles and forward inclination angles through numerical simulations.

MBE-ARI: A Multimodal Dataset Mapping Bi-Directional Engagement in Animal-Robot Interaction

Ian Noronha, Upinder Kaur

Pose EstimationMultimodalityBenchmark

🎯 What it does: Created the MBE-ARI multi-modal dataset and proposed a full-body pose estimation model for quadrupeds.

MDC-Seg: Multi-Directional Convolution-Based Semantic Segmentation for LiDAR Point Clouds

Ouyang Xin, Wei Liu

SegmentationAutonomous DrivingConvolutional Neural NetworkPoint Cloud

🎯 What it does: Propose MDC-Seg, which utilizes multi-directional convolution (MDConv) to perform parallel sparse feature encoding on bird's eye view (BEV) and range view (RV) planes, combined with attention mechanisms, effective multi-feature fusion (EMFF) module, and point voxel constraint (PVC) module, achieving effective receptive field expansion and accuracy improvement for 3D point cloud semantic segmentation.

ME-PATS: Mutually Enhancing Search-Based Planner and Learning-Based Agent for Tractor-Trailer Systems

Ke Fan, Zufeng Zhang

Autonomous Driving

🎯 What it does: Proposed the ME-PATS framework, aiming to plan dynamically feasible paths for tractor-trailer systems through mutually enhancing search planners and learning agents;

Measuring DNA Microswimmer Locomotion in Complex Flow Environments

T. Imamura, Sarah Bergbreiter

Optical FlowPhysics Related

🎯 What it does: This paper proposes a method using labeled microspheres to measure the motion of micro-swimming robots in complex flow environments.

Mechanisms and Computational Design of Multi-Modal End-Effector with Force Sensing Using Gated Networks

Yusuke Tanaka, Dennis W. Hong

Robotic IntelligenceRecurrent Neural Network

🎯 What it does: Designed and implemented a multi-modal end-effector called MAGPIE, capable of switching between flat-foot and line-foot modes and equipped with grasping functionality. It integrates eight-axis force sensing and Hall effect sensors to enable contact and tactile force measurement, and experimental validation demonstrates the performance of its foot functionality, sensing mechanism, ideal inverse model, and gated network model.

MEDiC: Autonomous Surgical Robotic Assistance to Maximizing Exposure for Dissection and Cautery

Xiao Liang, Michael C. Yip

OptimizationRobotic Intelligence

🎯 What it does: Developed the MEDiC system to maximize visual exposure and apply tissue tension during resection and hemostasis in autonomous surgical robots

MERLION: Marine ExploRation with Language guIded Online iNformative Visual Sampling and Enhancement

Shrutika Vishal Thengane, Malika Meghjani

RestorationVision Language ModelImageText

🎯 What it does: Proposes the MERLION framework for semantic alignment and visual enhancement in monitoring and exploration of turbid waters;

MeshDMP: Motion Planning on Discrete Manifolds Using Dynamic Movement Primitives

Matteo Dalle Vedove, Matteo Saveriano

Robotic IntelligenceMesh

🎯 What it does: Proposes the MeshDMP method, which uses learned demonstrations to enable the robot manipulator to embed workpiece geometric information extracted from triangular meshes through differential operators on a discrete manifold, and extends the Dynamic Movement Primitives (DMPs) framework to generate motion on mesh surfaces.

METDrive: Multimodal End-to-End Autonomous Driving with Temporal Guidance

Ziang Guo, D. Tsetserukou

Autonomous DrivingMultimodalityTime SeriesBenchmark

🎯 What it does: Designed and implemented a multi-modal end-to-end autonomous driving system METDrive, which utilizes embedded ego state time series features (rotation angle, steering, throttle signal, and waypoint vectors) for time-guided prediction of waypoints; simultaneously combines geometric features extracted from perceptual sensor data to jointly guide waypoint prediction.

MFSeg: Efficient Multi-Frame 3D Semantic Segmentation

Chengjie Huang, Krzysztof Czarnecki

SegmentationAutonomous DrivingPoint Cloud

🎯 What it does: Proposes MFSeg, an efficient multi-frame 3D semantic segmentation framework that aggregates point cloud sequences at the feature level and regularizes the feature extraction and aggregation processes.

MGSO: Monocular Real-Time Photometric SLAM with Efficient 3D Gaussian Splatting

Y. Hu, John S. Zelek

OptimizationComputational EfficiencyGaussian SplattingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Designed a monocular real-time photometric SLAM system called MGSO, integrating 3D Gaussian Splatting to achieve efficient dense 3D reconstruction.

MI-HGNN: Morphology-Informed Heterogeneous Graph Neural Network for Legged Robot Contact Perception

D. Butterfield, Lu Gan

Robotic IntelligenceGraph Neural NetworkGraph

🎯 What it does: Proposed a morphology-based heterogeneous graph neural network (MI-HGNN) for contact perception in legged robots.

MicroASV: An Affordable 3D-Printed Centimeter-Scale Autonomous Surface Vehicle

Kevin Macauley, Wei Wang

Robotic IntelligenceSimultaneous Localization and MappingImage

🎯 What it does: Designed, manufactured, and implemented MicroASV, a low-cost, centimeter-level autonomous surface vehicle.

MILE: Model-Based Intervention Learning

Yigit Korkmaz, Erdem Biyik

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Proposes a model-based intervention learning method called MILE, which learns control strategies by utilizing a small number of expert interventions along with feedback information during and outside interventions.

Milli-Scale AcousTac Sensing Using Soft Helmholtz Resonators

Jadesola Aderibigbe, Hannah Stuart

Physics RelatedAudio

🎯 What it does: The study utilizes pressure-driven soft Helmholtz resonators to achieve wireless tactile sensing, explores the relationship between emission frequency and resonance cavity geometry, and verifies that frequency increases when the resonance cavity is compressed can be measured by external microphones.

Miniature Dielectric Elastomer Actuator Probe Inspecting Confined Spaces Embedding a CMOS Sensor

Sahib Sandhu, Mihai Duduta

Robotic IntelligenceImage

🎯 What it does: Demonstrates a modular 2-DOF micro dielectric elastomer actuator (DEA) probe with an embedded CMOS sensor for visual data acquisition and navigation in confined spaces; verifies its ability to move through various complex confined paths.

Minimally Invasive Endotracheal Inside-Out Flexible Needle Driving System Towards Microendoscope-Guided Robotic Tracheostomy

Botao Lin, Hongliang Ren

Robotic IntelligenceBiomedical Data

🎯 What it does: Proposed an insertable flexible needle-driven system guided by a micro-endoscope and optical coherence tomography to perform robotic bronchoscopy through minimally invasive trans-tracheal puncture.

MiniVLN: Efficient Vision-and-Language Navigation by Progressive Knowledge Distillation

Junyou Zhu, Jing Liu

Computational EfficiencyKnowledge DistillationVision Language ModelMultimodality

🎯 What it does: Proposes a two-stage knowledge distillation framework to generate a lightweight Vision-and-Language Navigation model called MiniVLN.

Mitigating Side Effects in Multi-Agent Systems Using Blame Assignment

P. Rustagi, Sandhya Saisubramanian

OptimizationReinforcement Learning

🎯 What it does: Proposes a scheme that decomposes joint negative consequences (NSE) penalties into individual penalties through credit assignment, enabling decentralized multi-agent systems to alleviate NSE in shared environments.

Mixing Data-Driven and Geometric Models for Satellite Docking Port State Estimation Using an Rgb or Event Camera

C. Gentil, Teresa Vidal-Calleja

Pose EstimationData-Centric LearningImage

🎯 What it does: A lightweight data pipeline is proposed for satellite docking port detection and state estimation using monocular RGB or event cameras, leveraging shallow data-driven preprocessing to emphasize LM-MAP's reflective navigation assistance combined with geometric models for state estimation.

MJPR: Multi-Modal Joint Predictive Representation in Deep Reinforcement Learning

Zehan Wang, Haobin Shi

Representation LearningReinforcement LearningContrastive LearningMultimodality

🎯 What it does: Propose a multi-modal joint predictive representation (MJPR) method that predicts future latent states using multi-modal interactive information to improve sample efficiency in deep reinforcement learning.

mmDEAR: mmWave Point Cloud Density Enhancement for Accurate Human Body Reconstruction

Jiarui Yang, Ling Pei

Pose EstimationSuper ResolutionSafty and PrivacyPoint Cloud

🎯 What it does: Proposes a two-stage deep learning framework to enhance millimeter-wave point cloud density and improve human reconstruction accuracy.

Mobile-TeleVision: Predictive Motion Priors for Humanoid Whole-Body Control

Chenhao Lu, Xiaolong Wang

Robotic IntelligenceReinforcement LearningAuto Encoder

🎯 What it does: By separating upper body control from lower body movement, precise upper limb operations are achieved using inverse kinematics and motion repositioning, while robust lower limb gait control is implemented via reinforcement learning (RL), with Predictive Motion Priors (PMP) introduced to represent upper body motion.

MochiSwarm: A Testbed for Robotic Micro-Blimps in Realistic Environments

Jiawei Xu, David Saldaña

Robotic Intelligence

🎯 What it does: This paper introduces MochiSwarm, an open-source lightweight micro-robot airship test platform that employs modular hardware, a flexible software framework, and detachable perception modules to enable collaborative operations among multiple robots without relying on external localization systems. It demonstrates autonomous capabilities through a differential drive module combined with visual servoing, ultimately showcasing its autonomous performance in cargo pickup and delivery tasks involving up to 12 airships.

Model Predictive Control with Visibility Graphs for Humanoid Path Planning and Tracking Against Adversarial Opponents

Ruochen Hou, Dennis W. Hong

OptimizationRobotic IntelligenceGraph

🎯 What it does: This paper proposes and implements an obstacle avoidance, path planning, and trajectory tracking method based on Dynamic Augmented Visibility Graph (DAVG) and conflict-agnostic Model Predictive Control (cf-MPC) in the RoboCup 2024 adult robot soccer competition, enabling the robot to remain undefeated and win the championship in all matches.

Model Q-II: An Underactuated Hand with Enhanced Grasping Modes and Primitives for Dexterous Manipulation

Yinkai Dong, A. Dollar

Robotic Intelligence

🎯 What it does: Proposed Model Q-II, an undriven robotic hand that enhances multifunctional grasping by expanding grasp modes and manipulation primitives.

Model-Based Control Strategies Comparison of One Bionic Ankle Tensegrity Exoskeleton: BATE

Dunwen Wei, F. Ficuciello

Robotic Intelligence

🎯 What it does: Comparison of three model-based control strategies for the BATE exoskeleton

Model-Based Robotic Cell Aspiration: Tackling the Impact of Air Segment

Jiachun Zheng, Zhuoran Zhang

Robotic IntelligenceOrdinary Differential Equation

🎯 What it does: Construct a nonlinear dynamic model to reveal the effect of air segments inside the pipette on cell movement, and design a controller based on this model to achieve precise aspiration of human sperm

Model-Free Safety Filter for Soft Robots: A Q-Learning Approach

Guo Ning Sue, Guanya Shi

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose a plug-in, model-free safe filtering framework that uses Q-learning to filter potentially unsafe actions in any task-specific baseline policy.

Modeling and Control of Aerial Robot SERPENT: A Soft Structure Incorporated Multirotor Aerial Robot Capable of In-Flight Flexible Deformation

Shotaro Itahara, Ko Yamamoto

Robotic Intelligence

🎯 What it does: Proposed a control method for multirotor drones based on passive flexible components connection.

Modeling of Deformable Linear Objects Under Incomplete State Information

Marc Kilian Klankers, Jochen J. Steil

Pose EstimationRobotic Intelligence

🎯 What it does: Propose a fused network architecture to achieve state estimation and prediction of DLO under incomplete perception information by utilizing torque measurements.

Modeling Trust Dynamics in Robot-Assisted Delivery: Impact of Trust Repair Strategies

Dong Hae Mangalindan, Vaibhav Srivastava

Explainability and InterpretabilityRobotic IntelligenceSequential

🎯 What it does: In a robot-assisted delivery scenario, an Input-Output Hidden Markov Model (IOHMM) is constructed using human participant experimental data to capture dynamic changes in human trust toward robots, and to evaluate the impact of different trust repair strategies (short/long explanations, apology and commitment, denial) on trust recovery and maintenance.

Modeling Uncertainty in 3D Gaussian Splatting Through Continuous Semantic Splatting

Joey Wilson, Arnie Sen

SegmentationGaussian Splatting

🎯 What it does: Propose an algorithm for probabilistic updating and rasterization of semantic maps within the 3D Gaussian Splatting (3D-GS) framework.

MonLog: MONotonic-Constrained LOGistic Regressions for Automated Safety Curve Design

Alessandro Melone, Sami Haddadin

Biomedical Data

🎯 What it does: Propose the MonLog method, which utilizes a data-driven probabilistic model to automatically generate safety curves from injury protection data

MonoCT: Overcoming Monocular 3D Detection Domain Shift with Consistent Teacher Models

Johannes Meier, Daniel Cremers

Object DetectionDomain Adaptation

🎯 What it does: Propose an unsupervised domain adaptation method called MonoCT for monocular 3D object detection, generating high-quality pseudo labels for self-supervised learning.

Monocular Depth Estimation and Segmentation for Transparent Object with Iterative Semantic and Geometric Fusion

Jiangyuan Liu, Wei Zou

SegmentationDepth EstimationImage

🎯 What it does: Proposed a monocular framework that simultaneously performs segmentation and depth estimation of transparent objects using a single RGB image.

Monocular Visual Place Recognition in LiDAR Maps via Cross-Modal State Space Model and Multi-View Matching

Gongxin Yao, Yu Pan

RetrievalAutonomous DrivingComputational EfficiencyContrastive LearningSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose an efficient framework that utilizes a monocular camera to achieve cross-modal visual-point cloud retrieval in a pre-built LiDAR map, enabling monocular localization and reducing the computational burden of visual SLAM.

MonoDiff9D: Monocular Category-Level 9D Object Pose Estimation via Diffusion Model

Jian Liu, A. Mian

Pose EstimationDepth EstimationTransformerDiffusion modelContrastive LearningImagePoint Cloud

🎯 What it does: Proposes a monocular category-level 9D object pose estimation method called MonoDiff9D based on diffusion models, utilizing zero-shot depth estimation and point cloud feature fusion, and restoring the pose through a transformer denoiser.

MonoLDP: LED Assisted Indoor Mobile Bot Monocular Depth Prediction and Pose Estimation System

Chenxin Liang, Wenbo Ding

Pose EstimationDepth EstimationRobotic IntelligenceImage

🎯 What it does: Built and evaluated a monocular camera-based indoor mobile robot system called MonoLDP, integrating depth estimation, pose estimation, and visible light communication functions

Monotone Subsystem Decomposition for Efficient Multi-Objective Robot Design

Andrew Wilhelm, Nils Napp

OptimizationRobotic Intelligence

🎯 What it does: Developed a monotonic subsystem decomposition technique for efficiently computing the Pareto front of large-scale problems in multi-objective robot design.

MORDA: A Synthetic Dataset to Facilitate Adaptation of Object Detectors to Unseen Real-Target Domain While Preserving Performance on Real-Source Domain

Hojun Lim, Hyeongseok Jeon

Object DetectionData SynthesisDomain AdaptationImagePoint Cloud

🎯 What it does: Construct a synthetic fused domain MORDA, and jointly train 2D/3D object detection models using nuScenes and MORDA, evaluating on an unseen Korean real-world dataset AI-Hub.

MoRE: Unlocking Scalability in Reinforcement Learning for Quadruped Vision-Language-Action Models

Han Zhao, Zongyuan Ge

Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningMixture of ExpertsVision-Language-Action ModelMultimodality

🎯 What it does: This paper proposes a hybrid robot expert model called MoRE, which fine-tunes a large-scale vision-language-action model on mixed-quality data using reinforcement learning to achieve flexible execution of quadruped robots across various tasks.

MORF: Magnetic Origami Reprogramming and Folding System for Repeatably Reconfigurable Structures with Fold Angle Control

Gabriel Unger, Cynthia Sung

Physics Related

🎯 What it does: Proposes a magnetically reprogrammable and foldable MORF system that achieves a structure capable of repeated deformation while maintaining rigidity.

Motion Forecasting via Model-Based Risk Minimization

Aron Distelzweig, A. Valada

Autonomous DrivingOptimizationImageMultimodalityPoint Cloud

🎯 What it does: Proposed a multi-model trajectory prediction sampling method, generating optimal trajectories through a risk minimization framework.

Motion Tracks: A Unified Representation for Human-Robot Transfer in Few-Shot Imitation Learning

Juntao Ren, Jeannette Bohg

Domain AdaptationRepresentation LearningRobotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningVideoSequential

🎯 What it does: Propose a unified representation that expresses actions as short-term 2D trajectories, and build Motion Track Policy (MT-π), which can complete everyday tasks using only a small amount of human videos and limited robot demonstrations.

Motion-Guided Dual-Camera Tracker for Endoscope Tracking and Motion Analysis in a Mechanical Gastric Simulator

Yuelin Zhang, S. Cheng

Object TrackingTransformerBiomedical Data

🎯 What it does: Propose a dual-camera motion-guided visual tracker for 3D localization and motion analysis of the flexible endoscope tip in a mechanical stomach simulator.

MotionGlot: A Multi-Embodied Motion Generation Model

Sudarshan S. Harithas, Srinath Sridhar

GenerationRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningText

🎯 What it does: Propose the MotionGlot model, which can generate motions on various entities (such as quadruped robots and human bodies), achieved by adopting training methods of large language models (LLMs) and instruction tuning templates tailored for motion-related tasks.

MOVE: Multi-Skill Omnidirectional Legged Locomotion With Limited View in 3D Environments

Songbo Li, Qiuguo Zhu

Robotic IntelligenceContrastive Learning

🎯 What it does: Proposed the MOVE framework to achieve low-cost quadruped robots with multi-skilled omnidirectional gaits under limited perspectives.

MPC-QP-Based Control Framework for Compliant Behavior of Humanoid Robots in Physical Collaboration with Humans

Shubham S. Kumbhar, Panagiotis K. Artemiadis

OptimizationRobotic Intelligence

🎯 What it does: Proposed a control framework enabling humanoid robots to exhibit expected compliance when collaborating with humans to move heavy objects.

MPI-Mamba: Cross Propagation Mamba for Multipath Interference Correction

Kang An, Jindong Tian

Convolutional Neural NetworkTransformer

🎯 What it does: Studied multi-path interference correction methods, proposed a cross-propagation network based on Mamba, and designed an efficient and accurate real data collection scheme.

Ms. NAMI: Multimodal Semantic Navigation on Relative Metric Intention Graph

Shichao Zhai, Yue Wang

Autonomous DrivingGraph Neural NetworkReinforcement LearningMultimodality

🎯 What it does: Proposed the Ms. NAMI framework, which integrates multi-modal navigation tasks and achieves a unified navigation strategy based on a relative topological-metric intent graph.

MT-PCR: Leveraging Modality Transformation for Large-Scale Point Cloud Registration with Limited Overlap

Yilong Wu, Lu Zhang

Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose the MT-PCR method, achieving large-scale low-overlap point cloud registration using BEV conversion and 2D keypoint matching.

Multi-Agent Collective Construction of General Modular Structures

Irina Kostitsyna, Kenneth C. Cheung

OptimizationRobotic Intelligence

🎯 What it does: Proposes an algorithmic framework for multi-robot modular assembly systems, focusing on the NASA AR-MADAS architecture, and realizes the construction of generic module structures with non-histogram shapes.

Multi-Agent Ergodic Exploration Under Smoke-Based Time-Varying Sensor Visibility Constraints

Elena Wittemyer, H. Choset

Optimization

🎯 What it does: Studies multi-robot information acquisition path planning in time-varying smoke environments, achieving ergodic exploration based on a time-varying visibility model.

Multi-Agent Inverse Q-Learning from Demonstrations

Nathaniel Haynam, Erdem Biyik

Reinforcement Learning

🎯 What it does: Proposed a multi-agent inverse Q-learning framework called Multi-Agent Marginal Q-Learning from Demonstrations (MAMQL) for inferring reward functions from demonstrations.

Multi-Agent Obstacle Avoidance Using Velocity Obstacles and Control Barrier Functions

Alejandro Sánchez-Roncero, Petter Ögren

Autonomous DrivingRobotic Intelligence

🎯 What it does: Propose a multi-agent obstacle avoidance method combining Velocity Obstacles and Control Barrier Functions.

Multi-Agent Path Finding Using Conflict-Based Search and Structural-Semantic Topometric Maps

Scott Fredriksson, G. Nikolakopoulos

OptimizationRobotic IntelligenceGraphBenchmark

🎯 What it does: Propose using conflict-based search on sparse structural semantic topological maps for multi-robot path planning to accelerate computation and reduce conflict resolution frequency.

Multi-Agent Path Planning in Complex Environments using Gaussian Belief Propagation with Global Path Finding

Jens Hoigaard Jensen, Andriy Sarabakha

Autonomous DrivingOptimization

🎯 What it does: Proposes a method that combines Gaussian belief propagation with path integral and incorporates a new tracking factor for multi-agent path planning in complex environments, validated on RRT and lane structure-based planners.

Multi-Covering a Point Set by $m$ Disks with Minimum Total Area

Mariem Guitouni, Aaron Becker

Optimization

🎯 What it does: Study the problem of minimizing the total area for multiple coverage point sets, propose and analyze a fast heuristic algorithm, use it to initialize an exact integer programming solution, and enforce separation constraints between sensors by modifying the integer programming formulation and candidate circle sets.

Multi-Drone-Truck Collaborative Delivery with En Route Operations: A Hierarchical MARL-Based Approach

Shunong Hu, Rongqing Zhang

Autonomous DrivingReinforcement Learning

🎯 What it does: Developed a hierarchical multi-agent reinforcement learning-based method for multi-drone-unmanned vehicle collaborative delivery, supporting drone deployment and recovery during vehicle movement while considering drone path conflicts;

Multi-Floor Zero-Shot Object Navigation Policy

Lingfeng Zhang, Renjing Xu

Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelMultimodality

🎯 What it does: Proposed and implemented a Multi-Floor Zero-Shot Object Navigation Policy (MFNP), achieving cross-floor exploration and localization through three major components.

Multi-Goal Motion Memory

Yuanjie Lu, Xuesu Xiao

OptimizationRobotic Intelligence

🎯 What it does: Developed a multi-object motion memory technology to accelerate multi-target path planning for autonomous mobile robots in dynamic environments

Multi-Heuristic Robotic Bin Packing of Regular and Irregular Objects

Tim Nickel, K. Arras

OptimizationRobotic Intelligence

🎯 What it does: Proposes So-Pack, a general packing heuristic for irregularly shaped objects, and integrates it into a flexible weighted multi-heuristic planning system for robotic bin packing.

Multi-Layer Feature Exchange Transformer for Multi-View 6D Object Pose Estimation in Robot Bin Picking

Momen Khalil, Slobodan Ilic

Pose EstimationRobotic IntelligenceTransformerImage

🎯 What it does: Propose a Feature Exchange Transformer (FET) for early feature fusion in multi-view 6D pose estimation.

Multi-Layered Safety of Redundant Robot Manipulators Via Task-Oriented Planning and Control

Xinyu Jia, Haoyong Yu

Robotic Intelligence

🎯 What it does: Proposes a task-oriented planning and control framework for redundant robotic manipulators to achieve multi-layer safety while maintaining task execution efficiency;

Multi-Modality Test-Time Adaptation for Semantic Segmentation in Robotic Perception

Yan Liu, Yulan Guo

SegmentationDomain AdaptationRobotic IntelligenceMultimodality

🎯 What it does: Proposes the DMATA method, which includes a Momentum-based Teacher-Student framework, Uncertainty-Guide module, and 3D-Guide-2D fusion module, for multi-modal test-time adaptation in semantic segmentation.

Multi-Nonholonomic Robot Object Transportation with Obstacle Crossing Using a Deformable Sheet

Weijian Zhang, Masoumeh Mansouri

OptimizationRobotic Intelligence

🎯 What it does: Propose a two-phase iterative trajectory optimization framework for formation planning of multiple non-holonomic robots using deformable membranes in irregular and crowded environments, with special consideration of scenarios where robots cross obstacles across different topological classes.

Multi-Robot Collaboration Through Reinforcement Learning and Abstract Simulation

Adam Labiosa, Josiah P. Hanna

Robotic IntelligenceReinforcement Learning

🎯 What it does: Investigate the availability of abstract simulators in multi-agent reinforcement learning and transfer the trained strategies to real robot teams, focusing on cooperative robot soccer tasks.

Multi-Scale Convolutional Networks with Class-Normalized Logit Clipping for Robust Sea State Estimation from Noisy Ship Motion Data

Xin Qin, Shengyong Chen

ClassificationConvolutional Neural NetworkTime Series

🎯 What it does: Proposed an end-to-end neural network model that combines a feature extraction module based on deep representation learning with a specific loss function designed for noisy labels, enabling robust sea state estimation using ship motion data.

Multi-Segment Soft Robot Control Via Deep Koopman-Based Model Predictive Control

Lei Lv, Yu Luo

Robotic Intelligence

🎯 What it does: Proposes a Deep Koopman Model Predictive Control (DK-MPC) framework for controlling multi-segment soft robots

Multi-Task Invariant Representation Imitation Learning for Autonomous Driving

Jinghan Peng, Dehui Du

Autonomous Driving

🎯 What it does: A multi-task invariant representation imitation learning (MIRIL) method for autonomous driving is studied, combining invariant learning and imitation learning to extract cross-environment causal invariant representations from multi-scenario driving examples, and achieving policy learning, perception prediction, transfer dynamics learning, and other downstream branches through multi-task learning.

Multi-Task Robustness Enhancement Framework against Various Adversarial Patches

Lihua Jing, Xingxing Wei

Adversarial AttackImage

🎯 What it does: Proposes a unified robustness enhancement framework for various adversarial patches, which utilizes self-supervised learning to accurately locate different adversarial patches and employs an efficient adaptive patch filling method to mitigate their impact while maintaining visual coherence.

Multi-Type Preference Learning: Empowering Preference-Based Reinforcement Learning with Equal Preferences

Ziang Liu, Liang He

Reinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Proposes a multi-type preference learning (MTPL) method that simultaneously utilizes equal preference and explicit preference for training in preference reinforcement learning.

Multi-View Stereo with Geometric Encoding for Dense Scene Reconstruction

Guidong Yang (Chinese University Of Hong Kong), Ben M. Chen (Chinese University Of Hong Kong)

Depth EstimationImagePoint Cloud

🎯 What it does: Propose the GE-MVS multi-view stereo network, using geometric encoding to achieve more accurate and complete depth estimation and point cloud reconstruction.

Multimodal Behaviour Trees for Robotic Laboratory Task Automation

Hatem Fakhruldeen, A. I. Cooper

Robotic IntelligenceMultimodality

🎯 What it does: Propose a multi-modal perception-based behavior tree method for automating laboratory robot tasks (sample vial capping and lab rack insertion) and verifying task execution success.

Multiple Rotation Averaging with Constrained Reweighting Deep Matrix Factorization

Shiqi Li, Dingkun Wang

Optimization

🎯 What it does: Propose an unlabeled multi-rotation averaging method that directly solves in linear space by leveraging techniques such as deep matrix decomposition, low-rank symmetric neural networks, generative tree edge filtering, re-weighting, and dynamic depth selection.

Multirotor Target Tracking through Policy Iteration for Visual Servoing

Sotirios N. Aspragkathos, Kostas J. Kyriakopoulos

Object TrackingRobotic IntelligenceReinforcement LearningImage

🎯 What it does: A vision-based UAV flexible contour target tracking method is proposed by combining image moment descriptors with a policy iteration scheme.

MultiTalk: Introspective and Extrospective Dialogue for Human-Environment-LLM Alignment

Venkata Naren Devarakonda, F. Khorrami

Robotic IntelligenceTransformerLarge Language ModelAgentic AI

🎯 What it does: Propose MultiTalk, an LLM-based task planning method that achieves alignment between humans, environments, and LLMs through introspective and extrospective dialogue loops, generating plans consistent with environmental and execution agent capabilities.

Music-Driven Legged Robots: Synchronized Walking to Rhythmic Beats

Tai-Wei Hou, Lihua Zhang

Robotic IntelligenceAudio

🎯 What it does: Achieved gait control for quadruped robots based on music beat synchronization, proposing a hierarchical architecture that integrates low-level phase tracker, oscillator, and high-level phase modulator, enabling the robot to naturally synchronize with external music rhythms;

MuST: Multi-Head Skill Transformer for Long-Horizon Dexterous Manipulation with Skill Progress

Kai Gao, Jane Shi

Robotic IntelligenceTransformer

🎯 What it does: Propose a multi-head skill transformer (MuST) framework to learn and sequentially connect multiple motion primitives, enabling long-term dexterous manipulation tasks.

MVCTrack: Boosting 3D Point Cloud Tracking via Multimodal-Guided Virtual Cues

Zhaofeng Hu, Zhihang Yuan

Object TrackingAutonomous DrivingImageMultimodalityPoint Cloud

🎯 What it does: Proposed the Multimodal-Guided Virtual Clue Projection (MVCP) scheme to generate virtual clues for enriching sparse point clouds, and developed the enhanced tracker MVCTrack based on this.

Na Vid-4D: Unleashing Spatial Intelligence in Egocentric RGB-D Videos for Vision-and-Language Navigation

Haoran Liu, He Wang

Depth EstimationRobotic IntelligenceVision Language ModelVideo

🎯 What it does: Proposes NaVid-4D, a navigation agent based on a vision-language model (VLM), which relies solely on first-person RGB-D video streams for spatial understanding and reasoning to achieve precise instruction-following robotic actions.

Narrow Passage Path Planning Using Collision Constraint Interpolation

Minji Lee, Dongjun Lee

Optimization

🎯 What it does: Propose a framework that generates subproblems through continuous interpolation of collision constraints to maintain feasibility during narrow passage path planning.

NaviDiffusor: Cost-Guided Diffusion Model for Visual Navigation

Yiming Zeng, Hui Cheng

Autonomous DrivingDiffusion modelImage

🎯 What it does: Propose an RGB visual navigation framework that integrates learning and classical methods, first training a conditional diffusion model and then using differentiable cost gradients to guide path generation.

NavigateDiff: Visual Predictors are Zero-Shot Navigation Assistants

Yiran Qin, Ruimao Zhang

Autonomous DrivingTransformerLarge Language ModelVision Language ModelDiffusion model

🎯 What it does: A visual predictor was constructed by combining a large vision-language model with a diffusion network, enabling continuous prediction of the agent's next possible observations and embedding the predictions into a goal-oriented strategy to achieve zero-shot navigation.

NDOB-Based Control of a UAV with Delta-Arm Considering Manipulator Dynamics

Hongming Chen, Ximin Lyu

Autonomous DrivingRobotic Intelligence

🎯 What it does: Proposed and implemented a composite control scheme combining nonlinear disturbance observer (NDOB) and high-pass filter to enhance the maneuverability of UAV Delta-arm and the precision of the end-effector.

Near Time-Optimal Hybrid Motion Planning for Timber Cranes

Marc-Philip Ecker, Wolfgang Kemmetmüller

OptimizationRobotic Intelligence

🎯 What it does: Propose a time-optimal, collision-free hybrid motion planning method for hydraulic-driven wood cranes with passive joints

Nearest-Neighbourless Asymptotically Optimal Motion Planning with Fully Connected Informed Trees (FCIT*)

Tyler S. Wilson, J. Gammell

OptimizationBenchmark

🎯 What it does: Proposed a fully connected, informative, nearest-neighbor-free, and incrementally optimal motion planning algorithm called FCIT*, which constructs and searches a fully connected graph by utilizing SIMD parallelism to reduce edge evaluation costs;

NeRF-Based Transparent Object Grasping Enhanced by Shape Priors

Yi Han, Gan Ma

Pose EstimationRobotic IntelligenceNeural Radiance Field

🎯 What it does: Utilize NeRF combined with shape priors to achieve 3D reconstruction of transparent objects, and perform scene-level grasping prediction and robot grasping experiments based on complete 3D information.

NeuGrasp: Generalizable Neural Surface Reconstruction with Background Priors for Material-Agnostic Object Grasp Detection

Qingyu Fan, Shuo Wang

Object DetectionRobotic IntelligenceTransformer

🎯 What it does: Proposes a neural surface reconstruction method called NeuGrasp, which utilizes background prior to achieve material-insensitive grasp detection, particularly suitable for scenarios involving transparent and mirror-like objects.

Neural $\mathcal{L}_{1}$ Adaptive Control of Vehicle Lateral Dynamics

Pratik Mukherjee, Volkan Isler

Autonomous Driving

🎯 What it does: Proposed and implemented a Neural-L1 adaptive controller (Neural-L1) to learn uncertainties in the lateral error dynamics of front-wheel Ackermann vehicles, ensuring system stability and robustness.

Neural Dynamics Augmented Diffusion Policy

Ruihai Wu, Yunzhu Li

Robotic IntelligenceReinforcement Learning from Human FeedbackDiffusion modelWorld Model

🎯 What it does: Propose a neural dynamics-enhanced imitation learning method, which trains a robust diffusion strategy using a small number of demonstrations in the local support region and reorganizes objects beyond this region using an offline-trained neural dynamics model.

Neural Encodings for Energy-Efficient Motion Planning

Jocelyn Zhao, T. Aamodt

Autonomous DrivingComputational Efficiency

🎯 What it does: This paper proposes a neural motion planner using binary encoded labels (BEL), replacing traditional regression networks with a set of binary classifiers.