arXivSub Start free trial

IROS 2025 Papers — Page 12

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1984 papers

LSTM-MHSA-Enhanced Deep Reinforcement Learning for Accurate Gait Control in Human Musculoskeletal Model

Shiyu Mao, Dunwen Wei

Recurrent Neural NetworkTransformerReinforcement LearningTime SeriesBiomedical Data

🎯 What it does: Proposes a muscle dynamics deep reinforcement learning control model based on LSTM and MHSA-enhanced PPO algorithm for precise simulation of diverse human gaits.

LSW-Net: A Spatio-temporal Self-Supervised Framework for 2D LiDAR-Based Environment Perception

Haojie Dai, Qi Chen

Autonomous DrivingRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningPoint Cloud

🎯 What it does: Proposes LSW-Net, a self-supervised framework for learning environmental perception features from raw 2D LiDAR point clouds, incorporating an LS-Encoder with local convolutional perception and global attention, along with an interpretable weight extraction module.

LTLCodeGen: Code Generation of Syntactically Correct Temporal Logic for Robot Task Planning

Behrad Rabiei, N. Atanasov

Robotic IntelligenceLarge Language ModelText

🎯 What it does: Using large language models to convert natural language instructions into linear temporal logic (LTL) formulas, and combining them with semantic occupancy maps through motion planning algorithms to generate collision-free robot paths that satisfy the instructions.

LuSeg: Efficient Negative and Positive Obstacles Segmentation via Contrast-Driven Multi-Modal Feature Fusion on the Lunar

Shuaifeng Jiao, Huimin Lu

SegmentationConvolutional Neural NetworkContrastive LearningImageMultimodality

🎯 What it does: Developed the lunar surface simulation system LESS and the LunarSeg dataset, and proposed the LuSeg two-stage segmentation network for positive and negative obstacle segmentation.

Lywal-X: A Novel Wheel-claw Quadruped Robot

Hao Shen, Xuan Xiao

Robotic Intelligence

🎯 What it does: Designed and implemented the wheel-claw quadruped robot Lywal-X with omni-directional movement and grasping capabilities, developed motion strategies for climbing and grasping modes, analyzed its mobility performance, and experimentally verified the object transportation capability in single-claw and dual-claw modes.

M-Predictive Spliner: Enabling Spatiotemporal Multi-Opponent Overtaking for Autonomous Racing

Nadine Imholz, Michele Magno

Autonomous Driving

🎯 What it does: Using Kalman Filter-based multi-opponent tracking combined with spatial and velocity Gaussian Process Regression (GPR), the future intentions of opponents in head-to-head racing scenarios are predicted to achieve more accurate overtaking decisions.

M2H: Multi-Task Learning with Efficient Window-Based Cross-Task Attention for Monocular Spatial Perception

U. Udugama, F. Nex

SegmentationDepth EstimationTransformerImage

🎯 What it does: Proposed and implemented the Multi-Mono-Hydra (M2H) multi-task learning framework for semantic segmentation, depth estimation, edge detection, and surface normal estimation from monocular images.

M2P2: A Multi-Modal Passive Perception Dataset for Off-Road Mobility in Extreme Low-Light Conditions

A. Datar, Xuesu Xiao (George Mason University)

Robotic IntelligenceSimultaneous Localization and MappingImageMultimodalityPoint CloudTime SeriesBenchmark

🎯 What it does: Constructed a multi-modal passive perception dataset named M2P2, which includes thermal imaging, event cameras, stereo RGB cameras, GPS, two IMUs, and high-resolution LiDAR for ground truth. Multi-sensor calibration was completed, collecting 10 hours of offline data covering 32 km, encompassing different lighting conditions (bright, low-light, no-light), terrains (paved, off-road, non-off-road), as well as robot odometry and action information.

M3D-skin: Multi-material 3D-printed Tactile Sensor with Hierarchical Infill Structures for Pressure Sensing

Shunnosuke Yoshimura, Kei Okada

Robotic Intelligence

🎯 What it does: Developed a multi-material 3D printed tactile sensor named M3D-skin, utilizing the infill pattern of a multi-material FDM printer to construct hierarchical flexible structures with variable resistance. The study measures the impact of structural modifications on sensing performance, demonstrating the fabrication and application of multiple sensors, including foot motion pattern measurement, integration into a robotic hand, and tactile-based robotic manipulation.

M3PO: Massively Multi-Task Model-Based Policy Optimization

Aditya Narendra, Aleksandr Panov

OptimizationReinforcement LearningWorld Model

🎯 What it does: Propose an expandable multi-task model-based policy optimization framework called M3PO, which integrates an implicit world model with a hybrid exploration strategy;

MAD-GS:3D Gaussian Splatting for Motion and Defocus Images in Robotic Vision

Tianle Zeng, Ziqi Zheng

RestorationRobotic IntelligenceGaussian SplattingImage

🎯 What it does: Propose the MAD-GS framework to address motion blur and defocus blur in robotic vision tasks, while optimizing imprecise camera poses;

MADI: Malicious Agent Detection and Isolation in Mixed Autonomy Traffic Systems

Wei Hao, Lijun Chen

Anomaly DetectionAutonomous DrivingSafty and Privacy

🎯 What it does: Proposed the MADI framework for detecting and isolating path order violations and strategic congestion generation behaviors by malicious agents in mixed autonomous traffic systems.

MAER-Nav: Bidirectional Motion Learning Through Mirror-Augmented Experience Replay for Robot Navigation

Shanze Wang, Wei Zhang

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose the MAER-Nav framework to achieve bidirectional motion learning, utilizing mirror-augmented experience replay and curriculum learning to generate synthetic reverse navigation experiences from successful trajectories.

Mag-Match: Magnetic Vector Field Features for Map Matching and Registration

William McDonald, Teresa Vidal-Calleja

Pose EstimationRobotic IntelligenceSimultaneous Localization and MappingPhysics Related

🎯 What it does: Proposes the Mag-Match method, which achieves matching and registration between different maps using magnetic vector field features.

MagicGel: A Novel Visual-Based Tactile Sensor Design with Magnetic Gel

Jianhua Shan, Bin Fang

Robotic IntelligenceMultimodality

🎯 What it does: Designed and developed a visual tactile sensor called MagicGel that integrates a magnetic sensing mechanism. It uses strong magnetic particle markers and Hall sensors to real-time capture magnetic field changes, enabling force estimation and non-contact state information perception for home electronics.

Magnetic Microswarms with Controlled Locomotion in Liquid and Air Environments

Zih-Rou Chen, Na Liu

Physics Related

🎯 What it does: Designed and characterized magnetic microgroups based on hydrogel (composed of agarose hydrogel and NdFeB magnetic microparticles), achieving stable monolayer structures and synchronous movement in liquid environments under high-frequency (10 Hz) rotating magnetic fields; transitioning from monolayer to three-dimensional structures in air environments via oscillating magnetic fields, enabling navigation in complex terrains and interaction with tissue surfaces; and demonstrating potential for targeted delivery and adaptive filling of gastric perforations in ex vivo gastric tissue models.

Magnetically Actuated Steerable Catheter with Redundant DoF for Cardiovascular Interventions

Hongzhe Liao, Xiaoming Liu

Robotic Intelligence

🎯 What it does: Propose a magnetically controlled manipulable multi-degree-of-freedom catheter system to enhance the accuracy and safety of vascular interventions

Make Your AUV Adaptive: An Environment-Aware Reinforcement Learning Framework For Underwater Tasks

Yimian Ding, Yi Li

Large Language ModelReinforcement Learning

🎯 What it does: Proposes an environment-aware reinforcement learning framework to enhance the adaptability and decision-making capabilities of AUVs in underwater tasks.

MALMM: Multi-Agent Large Language Models for Zero-Shot Robotic Manipulation

Harsh Singh, Ivan Laptev

Robotic IntelligenceLarge Language ModelAgentic AI

🎯 What it does: Proposed a multi-agent large language model framework called MALMM for zero-shot robot manipulation. The framework assigns planning to high-level planning agents, low-level control agents, and supervisory agents, and after each step, combines environmental observations to achieve failure handling and adaptive replanning.

Mamba Policy: Towards Efficient 3D Diffusion Policy with Hybrid Selective State Models

Jiahang Cao, Renjing Xu

Computational EfficiencyRobotic IntelligenceDiffusion model

🎯 What it does: Propose Mamba Policy, which replaces UNet with the Mamba model as the 3D sampling policy network, reducing parameters by over 80%, and combines Mamba with Attention via the XMamba Block for deep feature extraction.

MambaGCN: Synergistic Integration of Graph Convolutional Networks and State Space Models for Point Cloud Processing

Zhifeng Rao, Zhiyun Lin

Graph Neural NetworkPoint Cloud

🎯 What it does: Developed a MambaGCN that combines graph convolutional networks and state space models for point cloud processing

MambaMap: Online Vectorized HD Map Construction using State Space Model

Ruizi Yang, Jianke Zhu

Autonomous DrivingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose the MambaMap framework, achieving online vectorized high-definition map construction by fusing long-range temporal features in the state space.

MambaNUT: Nighttime UAV Tracking via Mamba-based Adaptive Curriculum Learning

You Wu, Shuiwang Li

Object TrackingComputational EfficiencyVideo

🎯 What it does: Proposed MambaNUT, a pure Mamba-based tracking framework specifically designed for nighttime UAV tracking;

MambaPlace: Text-to-Point-Cloud Cross-Modal Place Recognition with Attention Mamba Mechanisms

Tianyi Shang, Fanchen Kong

RecognitionVision Language ModelTextPoint Cloud

🎯 What it does: Propose the MambaPlace framework to achieve coarse-to-fine stages for cross-modal localization between text and point clouds;

MambaSFLNet: A Mamba-based Model for Low-Light Image Enhancement with Spatial and Frequency Features

Mingyu Liu, Alois Knoll

RestorationImage

🎯 What it does: Proposed a Mamba-based framework called MambaSFLNet for low-light image enhancement, integrating spatial and frequency features.

ManeuverGPT Agentic Control for Safe Autonomous Stunt Maneuvers

Shawn Azdam, Aliasghar Arab

Autonomous DrivingLarge Language ModelPrompt Engineering

🎯 What it does: Developed the ManeuverGPT framework, which utilizes LLM-driven agents to generate and execute high-dynamic stunt maneuvers (e.g., J-turn) in the CARLA simulation environment, and adjusts control parameters through an iterative prompting method to accomplish complex actions without retraining model weights.

ManiDP: Manipulability-Aware Diffusion Policy for Posture-Dependent Bimanual Manipulation

Zhuo Li, Fei Chen

OptimizationRobotic IntelligenceDiffusion model

🎯 What it does: Proposed a dual-arm manipulation strategy called ManiDP based on diffusion models, which can generate dual-arm trajectories that meet posture requirements and optimize dual-arm configurations to satisfy posture-dependent task demands.

ManiGaussian++: General Robotic Bimanual Manipulation with Hierarchical Gaussian World Model

Tengbo Yu, Ziwei Wang

Robotic IntelligenceGaussian SplattingWorld Model

🎯 What it does: Proposes a multi-task dual-arm manipulation method called ManiGaussian++, which models multi-body spatiotemporal dynamics using a hierarchical Gaussian world model.

Manip4Care: Robotic Manipulation of Human Limbs for Solving Assistive Tasks

Yubin Koh, A. H. Qureshi

Robotic Intelligence

🎯 What it does: Developed a modular simulation pipeline enabling robots to grasp and reposition human limbs to enhance assistive care functions.

ManipGPT: Is Affordance Segmentation by Large Vision Models Enough for Articulated Object Manipulation?

Taewhan Kim, Hao Dong

SegmentationRobotic IntelligenceTransformerSupervised Fine-TuningImage

🎯 What it does: Propose the ManipGPT framework, which leverages large-scale pre-trained ViT to predict optimal interaction regions for articulated objects, achieves component-level functional segmentation with minimal training, and completes grasping and manipulation in both simulated and real environments.

Manipulate-To-Navigate: Reinforcement Learning with Visual Affordances and Manipulability Priors

Yuying Zhang, J. Pajarinen

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: This paper proposes a mobile manipulation method based on reinforcement learning, which utilizes visual manipulability maps and manipulability priors to learn a strategy of manipulating obstacles before navigation in dynamic environments. The effectiveness is validated in Reach and Door simulation tasks on the Spot robot, ultimately achieving successful completion of the Reach task on a real Spot robot.

Manipulation of Elasto-Flexible Cables with Single or Multiple UAVs

C. Gabellieri, Antonio Franchi

Robotic IntelligencePhysics Related

🎯 What it does: Studied a system of multiple quadrotor drones manipulating a deformable, extendable rope, using a discretized model to divide the rope into linear springs and passive spherical joints with concentrated masses, and identified the system's planar output; verified the rope manipulation based on planar trajectory through numerical simulation, then experimentally validated the model's effectiveness on a two-robot experimental platform, and tested the model using a model-based closed-loop controller with rope output feedback.

Many-Objective Motion Generation Method for Redundant Manipulators by Solving Pathwise Inverse Kinematics

Bin Xie, Di Wu

OptimizationRobotic Intelligence

🎯 What it does: A multi-objective motion generation method for redundant manipulators is proposed to solve the path inverse kinematics problem.

MapDiffusion: Generative Diffusion for Vectorized Online HD Map Construction and Uncertainty Estimation in Autonomous Driving

T. Monninger, Sihao Ding

GenerationAutonomous DrivingDiffusion model

🎯 What it does: Propose MapDiffusion, which generates vectorized online HD maps using diffusion models and provides uncertainty estimation;

Mapless Collision-Free Flight via MPC using Dual KD-Trees in Cluttered Environments

Linzuo Zhang, Danping Zou

Autonomous DrivingOptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Propose a map-free obstacle avoidance method based on MPC and dual KD-Tree, directly generating safe actions using sparse path points and depth camera point clouds without requiring 3D map construction or trajectory tracking.

Mapping in Indoor Environments Including Transparent Objects Using Stereo Polarization Camera and Projector

Yusuke Ogihara, Atsushi Yamashita

Depth EstimationSimultaneous Localization and MappingImage

🎯 What it does: Propose a method to generate maps containing transparent objects in indoor environments using a stereo polarizing camera and projector, improving the signal-to-noise ratio (SNR) of polarization measurements and removing RGB information during polarization depth estimation to avoid depth estimation of objects behind glass; and generating complete environment maps by combining multi-position depth estimation with self-localization.

MarineGym: A High-Performance Reinforcement Learning Platform for Underwater Robotics

Shuguang Chu, Canjun Yang

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: Introduces MarineGym, a high-performance reinforcement learning platform designed for underwater robots.

Markov Parameters Generation for Data-based Modeling of Tensegrity Robots Considering Finite Word-Length Effects

Linxuan Shi, Yuling Shen

Robotic Intelligence

🎯 What it does: Study the impact of finite word length effects on Markov parameters of tension structure robots, and propose a method to generate accurate Markov parameters in digital simulations.

MARS-FTCP: Robust Fault-Tolerant Control and Agile Trajectory Planning for Modular Aerial Robot Systems

Rui Huang, Lin Zhao

OptimizationRobotic Intelligence

🎯 What it does: Proposes a fault-tolerant control reallocation method and flexible trajectory planning method for the modular aerial robotic system (MARS), achieving fault tolerance and collision avoidance flight under multi-module and different configurations.

MARSCalib: Multi-robot, Automatic, Robust, Spherical Target-based Extrinsic Calibration in Field and Extraterrestrial Environments

Seokhwan Jeong, Younggun Cho

SegmentationPose EstimationRobotic IntelligenceTransformerImagePoint Cloud

🎯 What it does: A multi-robot automatic robust extrinsic calibration method based on spherical targets is proposed, applicable to outdoor and extraterrestrial environments. It can accurately extract spherical and elliptical centers and solve the transformation matrix even when the target or sensors are damaged.

Masked Autoencoders are Robust Task Offloaders for Timely and Accurate Inference

Wonyeong Lee, Jinkyu Lee

Anomaly DetectionComputational EfficiencyRobotic IntelligenceAuto EncoderImage

🎯 What it does: Propose a collaborative edge server framework for image reconstruction and task offloading that balances computational efficiency and image processing accuracy, targeting edge devices in hazardous environments such as robots, enabling image anomaly detection tasks under resource-constrained and unstable network conditions.

MaskSem: Semantic-Guided Masking for Learning 3D Hybrid High-Order Motion Representation

Wei Wei, Jianqin Yin

RecognitionTransformerGraphSequential

🎯 What it does: Propose MaskSem, a semantic-guided occlusion method for learning 3D mixed high-order motion representations to enhance skeletal action recognition.

Mastering the Labyrinth Game: Efficient Multimodal Reinforcement Learning with Selective Reconstruction

Thomas Bi, Raffaello D'Andrea

Reinforcement LearningMultimodality

🎯 What it does: Achieving higher sample efficiency and fully autonomous learning in maze games through selective reconstruction, prioritized experience replay, and automatic pinball reloader

MATRICS: A Multi-Agent Deep Reinforcement Learning-Based Traffic-Aware Intelligent Lane-Change System

L. Das, Myounggyu Won

Autonomous DrivingReinforcement Learning

🎯 What it does: Designed and evaluated a traffic perception intelligent lane-changing system based on multi-agent deep reinforcement learning, capable of simultaneously optimizing traffic efficiency, driving safety, and ride comfort for both the ego vehicle and the overall road traffic.

MaxAuc: A Max-Plus-Based Auction Approach for Multi-Robot Allocations for Time-Ordered Temporal Logic Tasks

Mengjie Wei, Xiang Yin

OptimizationRobotic Intelligence

🎯 What it does: Propose the MaxAuc algorithm for task allocation in multi-robot systems operating in discrete workspaces to complete linear temporal logic tasks while considering temporal order constraints between tasks.

Maximum Clique-Based Floorplan Association for Robust Multi-Session Stereo SLAM in Challenging Indoor Environments

Haolin Wang, Yihong Wu

OptimizationRobotic IntelligenceSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a hierarchical multi-scenario SLAM framework that uses long-term plane maps as global features to achieve robust localization and map fusion under extreme viewpoints and lighting changes.

MCE-based Direct FTC Method for Dynamic Positioning of Underwater Vehicles with Thruster Redundancy

Ji-Hong Li

Robotic Intelligence

🎯 What it does: Proposes a model-driven, active fault-tolerant control (FTC) method based on motion control error (MCE) for dynamic positioning of underwater vehicles with thruster redundancy; the method directly utilizes MCE to construct residuals during steady-state operation of the control system, detecting thruster faults and achieving control reconfiguration.

MCTrack: A Unified 3D Multi-Object Tracking Framework for Autonomous Driving

Xiyang Wang, Mu Yang

Object TrackingAutonomous DrivingPoint Cloud

🎯 What it does: Proposed a unified 3D multi-object tracking framework called MCTrack, which achieves good performance on the KITTI, nuScenes, and Waymo datasets.

Measuring Uncertainty in Shape Completion to Improve Grasp Quality

Nuno Ferreira Duarte, José Santos-Victor

Robotic IntelligenceConvolutional Neural NetworkPoint Cloud

🎯 What it does: Proposes a method for calculating the uncertainty of 3D shape completion models during single-view point cloud inference, and uses this uncertainty to update the quality score of the grasping pose algorithm to improve grasping quality.

Mechanically Programming the Cross-Sectional Shape of Soft Growing Robotic Structures for Patient Transfer

O. G. Osele, Allison M. Okamura

Robotic Intelligence

🎯 What it does: Designed and verified a method utilizing flexible strips for mechanically programming the cross-sectional shape of a soft continuously growing robot structure, achieving a wide and thin cross-section during expansion and maintaining multi-axis flexibility during compression;

Mechanism Design, Optimization, and Experimental Validation of an Ultrasound-Guided Series-Parallel Hybrid Robot for Prostate Transperineal Puncture*

Haiyuan Li, Yilun Shi

OptimizationRobotic IntelligenceUltrasound

🎯 What it does: This paper designs and implements a series of parallel hybrid robots guided by ultrasound for prostate percutaneous puncture, and verifies its feasibility through prototype experiments.

MEFusion: Memory-Efficient Data Fusion for Real-Time 3D Reconstruction On Resource-Constrained Devices

Ruizhi Cao, Chenhao Xie

Computational EfficiencyPoint Cloud

🎯 What it does: Designed the MEFusion framework, replacing traditional element-wise histogram aggregation with probabilistic updates using a single instance label to achieve semantic fusion during real-time 3D reconstruction.

MeGS-SLAM:Memory Efficient Gaussian Splatting SLAM with Graph Signal Processing

Sude Zhang, Zhiyong Zhang

Gaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Propose a memory-efficient Gaussian Splatting SLAM framework (MeGS-SLAM), which improves edge reconstruction quality and significantly reduces the number of Gaussians by incorporating edge prior constraints and graph signal processing.

MelumiTac: Vision-based Tactile Sensor Using Mechanoluminescence for Dynamic Tactile and Nociceptive Perception

S. Bae, Kyungseo Park

Robotic IntelligenceOptical FlowImageVideo

🎯 What it does: Developed a visual tactile sensor named MelumiTac, which utilizes electro-mechanical luminescent materials to self-illuminate under dynamic tactile stimuli, enabling the visualization of dynamic tactile events and pain responses, as well as real-time deformation tracking.

MemGS: Memory-Efficient Gaussian Splatting for Real-Time SLAM

Yi-Feng Bai, Yi Zhou

Computational EfficiencyGaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Improved the GPU memory usage of 3D Gaussian Splatting (3DGS) in SLAM and enhanced rendering quality, primarily by merging redundant 3D Gaussian primitives based on geometric similarity in voxel space, and by initializing Gaussian primitives using Patch-Grid point sampling for more accurate scene modeling.

Memory-Efficient Real Time Many-Class 3D Metric-Semantic Mapping

Vallabh Nadgir, Kris Hauser

CompressionComputational EfficiencyAuto Encoder

🎯 What it does: This study proposes two schemes for compressing semantic fusion memory to enable real-time construction of multi-class 3D metric semantic maps.

Merry-Go-Round: Safe Control of Decentralized Multi-Robot Systems with Deadlock Prevention

Wonjong Lee, Changjoo Nam

Robotic Intelligence

🎯 What it does: Proposes a hybrid method for decentralized multi-robot navigation that ensures safety and prevents deadlocks;

meSch: Multi-Agent Energy-Aware Scheduling for Task Persistence

Kaleb Ben Naveed, Dimitra Panagou

OptimizationRobotic Intelligence

🎯 What it does: Developed an energy-aware scheduling protocol for multi-robot teams aimed at long-term persistent tasks

Mesh-Learner: Texturing Mesh with Spherical Harmonics

Yunfei Wan, Fu Zhang

GenerationPoint CloudMesh

🎯 What it does: Propose a 3D reconstruction and rendering framework named Mesh-Learner, which can directly integrate with traditional rasterization pipelines, learns perspective-related radiance for each mesh, and achieves end-to-end training through spherical harmonic (SH) texture mapping.

MetaFold: Language-Guided Multi-Category Garment Folding Framework via Trajectory Generation and Foundation Model

Haonan Chen, Lin Shao

Robotic IntelligenceLarge Language ModelVision-Language-Action ModelTextPoint Cloud

🎯 What it does: Propose the MetaFold framework, which decomposes the garment folding task into task planning and action prediction, implemented using language-guided point cloud trajectory generation and a low-level foundation model.

Method for Sensing Lateral Force and Skidding on the Tool Tip in Surgical Robot Deep Bone Drilling *

Zheyu Chen, Liang Li

Robotic Intelligence

🎯 What it does: Developed a bone drilling surgical robot with slip perception capability, and proposed a sensing scheme utilizing rigid body force transmission to achieve separation of thrust and lateral forces. Additionally, a tool tip slip estimation method was designed based on the bending curve model and spatial constraint model (SBCM), with a dedicated simulation device constructed for verification.

MGPRL: Distributed Multi-Gaussian Processes for Wi-Fi-based Multi-Robot Relative Localization in Large Indoor Environments

Sai Krishna Ghanta, Ramviyas Parasuraman

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposes the MGPRL framework, which utilizes Wi-Fi RSSI and convex hulls of multiple APs for distributed multi-robot relative localization. It employs co-located multi-output Gaussian processes for RSSI field prediction and combines uncertainty-aware multi-AP localization with weighted convex hull alignment to achieve robust relative pose estimation.

MIAT: Maneuver-Intention-Aware Transformer for Spatio-Temporal Trajectory Prediction

Chandra Raskoti, Weizi Li

Autonomous DrivingTransformerTime Series

🎯 What it does: Proposed and implemented the Maneuver-Intention-Aware Transformer (MIAT) architecture, combining maneuver-intention-aware control mechanisms with spatiotemporal interaction modeling to enhance vehicle trajectory prediction

Micro-robotic Swarm of Silicone Oil-based Ferrofluid’s Microdroplets

Yulei Fu, Wendong Wang

Robotic IntelligencePhysics Related

🎯 What it does: A micro-robot swarm system based on silicone oil-based ferromagnetic liquid microdroplets was developed, achieving three reconfigurable self-organizing modes (aggregation, dispersion, and chain-like structures) through 3D magnetic field control. Additionally, two motion modes (sliding and rolling) were discovered, with the sliding mode enabling navigation in narrow complex channels and directional bubble transport, accomplishing translational and rotational movements.

Micro-UAV with Ant-Inspired Bistable Gripper for Adaptive Perching and Wildlife Detection

Yuan Liu, Shimin Wei

🎯 What it does: Designed and verified a miniature quadrotor UAV equipped with a bio-inspired dual-stable gripper mimicking ants for adaptive perching and wildlife detection.

Microfluidics-Based Analysis of Controlled Mixing and Bubble Formation in Soda Solutions for Education

E. Owusu, Tao Yue

ImagePhysics Related

🎯 What it does: In classroom experiments, PDMS microfluidic chips, injection pumps, and high-resolution microscopes are used to observe the mixing and bubble generation of vinegar and baking soda solutions under laminar flow conditions.

Minimizing Acoustic Noise: Enhancing Quiet Locomotion for Quadruped Robots in Indoor Applications

Zhanxiang Cao, Yue Gao

OptimizationRobotic Intelligence

🎯 What it does: Optimize the acoustic noise of quadruped robots when walking in indoor environments

MinkOcc: Towards real-time label-efficient semantic occupancy prediction

Samuel Sze, L. Kunze

Autonomous DrivingComputational EfficiencyConvolutional Neural NetworkImageMultimodalityPoint Cloud

🎯 What it does: Proposes the MinkOcc framework, which utilizes multi-modal camera and LiDAR data, employing a two-step semi-supervised training method: first pre-training with a small amount of 3D annotated data, then continuing supervision through accumulated LiDAR sweeps and image annotations generated by a visual foundation model, achieving real-time semantic occupancy prediction.

MISCGrasp: Leveraging Multiple Integrated Scales and Contrastive Learning for Enhanced Volumetric Grasping

Qingyu Fan, Shuo Wang

Robotic IntelligenceTransformerContrastive Learning

🎯 What it does: Proposes MISCGrasp, a volumetric grasping method combining multi-scale feature extraction with contrastive learning, capable of adapting to objects of different shapes and sizes.

Mitigating Hallucinations in YOLO-based Object Detection Models: A Revisit to Out-of-Distribution Detection

Weicheng He, S. Bensalem

Object DetectionData SynthesisAutonomous DrivingConvolutional Neural NetworkSupervised Fine-TuningImageBenchmark

🎯 What it does: Investigate the hallucination problem in YOLO detection models, calibrate existing OoD benchmarks, and propose a joint detection-filtering pipeline. By fine-tuning with synthetically generated semantically similar OoD data, hallucination errors are reduced.

MIVG: Mode-Isolated Velocity-Guide Algorithm for Quadratic Optimization-Based Obstacle Avoidance

Hangyu Lin, Kunpeng Wu

OptimizationRobotic Intelligence

🎯 What it does: Proposed the Mode-Isolated Velocity-Guide (MIVG) algorithm, integrating a dual-mode isolation strategy with Velocity-Guide Potential Field (VGPF), addressing the issue of execution failure in dynamic obstacle avoidance caused by lack of directional guidance and constraint conflicts.

Mixed Integer Conic Programming for Multi-Agent Motion Planning in Continuous Space

Shizhe Zhao, Zhongqiang Ren

Optimization

🎯 What it does: Proposes a multi-agent motion planning method based on Mixed Integer Cone Programming (MICP) and validates its effectiveness through experiments.

MIXPINN: Mixed-Material Simulations by Physics-Informed Neural Network

Xintian Yuan, Philipp Fuernstahl

Computational EfficiencyGraph Neural NetworkMeshGraphPhysics Related

🎯 What it does: Developed a physics-informed graph neural network (MIXPINN) framework for simulating interactions between soft tissues and rigid structures, utilizing graph enhancement to achieve soft-rigid interaction modeling.

MK-Pose: Category-Level Object Pose Estimation via Multimodal-Based Keypoint Learning

Yifan Yang, Jingtai Liu

Pose EstimationGraph Neural NetworkImageTextMultimodalityPoint Cloud

🎯 What it does: A multi-modal keypoint learning framework, MK-Pose, is proposed to achieve category-level object pose estimation by integrating RGB images, point clouds, and category-level text descriptions.

MM-Geo: Multi-Scale and Multi-Positive UAV-View Geo-Localization

Pan Ai, Guoquan Huang

RetrievalContrastive LearningImage

🎯 What it does: Propose an MM-Geo method that uses uniformly sized satellite tiles and real-time retrieval of matching tiles from drone images at different altitudes, achieving geolocation from the drone's perspective.

MMCD: Multi-Modal Collaborative Decision-Making for Connected Autonomy with Knowledge Distillation

Rui Liu, Ming C. Lin

Autonomous DrivingKnowledge DistillationImageMultimodalityPoint Cloud

🎯 What it does: Propose the MMCD framework, which fuses RGB images and LiDAR point clouds from the owner's vehicle and collaborative vehicles through multi-modal fusion to enhance decision-making capabilities in challenging environments.

mmWave Radar-Based Non-Line-of-Sight Pedestrian Localization at T-Junctions Utilizing Road Layout Extraction via Camera

Byeonggyu Park, Seong-Woo Kim

Autonomous DrivingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose a framework that extracts road layout through a camera to interpret millimeter-wave radar point clouds, thereby locating pedestrians in non-line-of-sight (NLoS) areas.

MobiExo: GPS-SLAM Fusion for Seamless Indoor-Outdoor Mobile Manipulation with Hand-Foot Coordination

Jianpeng Wang, F. Yu

Federated LearningRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Developed the MobiExo system to achieve remote control for seamless indoor and outdoor mobile operations, integrating GPS and SLAM localization with hand-foot coordinated control.

Mobile Manipulator For Robotic Lacrosse: Learning to Pass the Ball

Xinchi Huang, Yi Guo

Robotic Intelligence

🎯 What it does: Developed a mobile manipulator for playing volleyball and achieved ball passing tasks between two robots

Model Predictive Control for 3D Steerable Needles: A Hierarchical Approach to Reduce Tissue Trauma

Sajjad Hussain, Fanny Ficuciello

OptimizationRobotic Intelligence

🎯 What it does: Proposes a three-dimensional control framework that integrates model predictive control (MPC) with hierarchical supervisory logic for the control of a bevel-tip steerable needle.

Model Predictive Control for Cable-Driven Remote Actuation Systems with Friction and Compliance

Moein Forouhar, Sami Haddadin

OptimizationRobotic Intelligence

🎯 What it does: Constructed and controlled a cable-driven remote execution system with friction and elastic properties.

Model-Based External Wrench Estimation for Underwater Robots

Moritz Graf, Daniel A. Duecker

Robotic IntelligencePhysics Related

🎯 What it does: Model-based external force estimation methods are extended from aerial drones to underwater robots, incorporating hydrodynamic effects and DVL sensor fusion.

Model-Driven Development of Distributed Controllers Using Petri Nets and Low-Code Strategy

Luis Gomes, João-Paulo Barros

🎯 What it does: Propose a model-driven distributed controller development method based on Petri nets, and achieve end-to-end support from modeling to deployment through a low-code strategy;

Model-Free Catheter Delivery Strategy for Robotic Transcatheter Tricuspid Valve Replacement

Haichuan Lin, Shuangyi Wang

OptimizationRobotic Intelligence

🎯 What it does: Proposed a model-free robotic catheter delivery strategy for transcatheter tricuspid valve replacement (TTVR).

Model-Mediated Teleoperation with 3D Dynamic Environment Tracking (MMT-DET): A Comparative Study of Task Performance with Time-Domain Passivity Control

Diego Fernandez Prado, Eckehard G. Steinbach

Robotic IntelligenceImage

🎯 What it does: Developed a vision-based MMT-DET system and conducted user studies

Modeling and Evaluating Trust Dynamics in Multi-Human Multi-Robot Task Allocation

Ike Obi, Byung-Cheol Min

OptimizationRobotic Intelligence

🎯 What it does: Propose the Expectation Confirmation Trust (ECT) model and evaluate its application in task allocation for multi-human-multi-robot team collaboration.

Modeling and Simulation of Single-micropipette Cell Rotation for Imitation Learning

Zefu Wang, Mingzhu Sun

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed a 3D simulation system for single microtubule cell rotation and implemented imitation learning-based control for cell rotation within this simulation environment;

Modeling Deception in Multi-Robot Target-Attacker-Defender Game via Deep Reinforcement Learning

Fandi Gou, Yunze Cai

Robotic IntelligenceReinforcement Learning

🎯 What it does: A hierarchical decision-making framework is proposed, using multi-agent reinforcement learning (MARL) for high-level deception strategy planning and optimal control for low-level motion control, with a composite deception-oriented reward function designed that includes hit rewards, belief-switching rewards, and positional advantage rewards; this framework is implemented to achieve deceptive behaviors in multi-robot target-attacker-defender (MR-TAD) games.

Modeling Human-like Driving Behavior Based on Maximum Entropy Deep Inverse Reinforcement Learning

Jiamin Shi, Jingmin Xin

Autonomous DrivingReinforcement LearningVideo

🎯 What it does: Propose a sampling method based on maximum entropy deep inverse reinforcement learning (MEDIRL) for simulating human driving behavior and evaluating the rewards of candidate trajectories

Modeling of Viscoelastic Liquid Crystal Elastomer Actuators

Hao Wang, Jian Zhu

Physics Related

🎯 What it does: Developed a dynamic model for liquid crystal elastomer (LCE) actuators under light stimulation, derived the relationship between laser power and temperature changes, and analyzed their viscoelastic behavior using a spring-dashpot framework. The dynamic equations for linear contraction actuators and bending actuators were validated with acceptable error.

Modeling The States of Liquid Phase Change Pouch Actuators by Reservoir Computing

Cedric Caremel, Tung D. Ta

Robotic IntelligenceRecurrent Neural NetworkTime SeriesPhysics Related

🎯 What it does: Proposed and implemented a reservoir computing-based modeling method for the inflation state of liquid-phase transition foam actuators, and designed a soft gripper with dual foam actuators.

Modular Decision-Making and Drivable Areas for Multi-Agent Autonomous Racing

A. Toschi, Marko Bertogna

Autonomous DrivingOptimization

🎯 What it does: Proposes an interactive perception modular framework for local trajectory planning in multi-agent racing scenarios; first, determines the drivable 'tunnel' area based on predictions of other vehicles' behaviors, then selects the optimal path between static and dynamic vehicles through a high-level decision module, and decides the timing for following or overtaking; subsequently, computes optimal, collision-free trajectories within the selected tunnel using an MPC module.

Modular Soft Wearable Glove for Real-Time Gesture Recognition and Dynamic 3D Shape Reconstruction

Huazhi Dong, Yunjie Yang

RecognitionPose EstimationConvolutional Neural NetworkTransformerPoint Cloud

🎯 What it does: Developed a high-sensitivity, modular flexible capacitive sensor glove based on linear electrodes and liquid metal (EGaIn), capable of independently capturing bending information of each finger joint and measuring subtle spacing between adjacent fingers, thereby achieving gesture recognition and dynamic 3D hand shape reconstruction;

MODUR: A Modular Dual-reconfigurable Robot

Jie Gu, Dan Zhang

Robotic Intelligence

🎯 What it does: Proposed a novel dual-reconfigurable robot MODUR, which can achieve high-level self-reconfiguration at the module level, and each module can perform basic movements through deformation.

MoE-Loco: Mixture of Experts for Multitask Locomotion

R. Huang, Hang Zhao

Robotic IntelligenceReinforcement LearningMixture of Experts

🎯 What it does: Proposes MoE-Loco, a mixture-of-experts framework for multi-task locomotion, enabling a single policy to navigate various terrains (e.g., beams, pits, stairs, slopes, and obstacles) as well as bipedal/quadrupedal gaits.

Monocular One-Shot Metric-Depth Alignment for RGB-Based Robot Grasping

Teng Guo, Jingjin Yu

Depth EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Propose a monocular single-sampling metric depth alignment framework (MOMA), which utilizes sparse depth points during camera calibration to align scale, rotation, and translation, thereby recovering accurate metric depth from a single RGB image and supporting fine-tuning for transparent objects;

Monocular Person Localization under Camera Ego-Motion

Yu Zhan, Hong Zhang

Pose EstimationImage

🎯 What it does: Propose a joint optimization method based on the four-point model to simultaneously estimate the camera pose and the 3D position of a person under a mobile monocular camera;

MORE: Mobile Manipulation Rearrangement Through Grounded Language Reasoning

Mohammad Mohammadi, A. Valada

Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelMultimodalityGraphBenchmark

🎯 What it does: Proposed a method called MORE that utilizes language models for zero-shot manipulation planning, focusing on rearrangement tasks.

Morphological Computation in Robotic Hopping: The Role of Monoarticular and Biarticular Muscle Configurations

Marc Murcia, Maziar Ahmad Sharbafi

Robotic Intelligence

🎯 What it does: Studied how lower limb joints recover during sudden ground drops using the bionic humanoid robot EPA-Hopper II, and explored the role of the single-joint muscle soleus (SOL) and the bi-articular muscle gastrocnemius (GAS) in the recovery process.

Motion Control of a Hybrid Self-Reconfigurable Wheel-Legged Dual-Arm Robot

Rui Zhang, Wenjie Song

Robotic Intelligence

🎯 What it does: Designed and verified a self-reconfigurable wheg-legged dual-arm robot, and proposed a distributed operational paradigm and a unified modular control architecture.