arXivSub Start free trial

ICRA 2025 Papers — Page 4

IEEE International Conference on Robotics and Automation · 1604 papers

CTS: A Consistency-Based Medical Image Segmentation Model

Kejia Zhang, Bao Yu

SegmentationBiomedical Data

🎯 What it does: Using a consistency model for segmenting medical images, and designing a multi-scale feature signal supervision pattern and loss function guidance to achieve model convergence.

CTSAC: Curriculum-Based Transformer Soft Actor-Critic for Goal-Oriented Robot Exploration

Chunyu Yang, Xin Zhang

Robotic IntelligenceTransformerReinforcement LearningPoint Cloud

🎯 What it does: Proposed a Transformer-based Soft Actor-Critic algorithm with Curriculum Learning (CTSAC), integrating Transformer into the SAC framework to leverage historical information, and employing periodic review-based curriculum learning to enhance training efficiency and prevent catastrophic forgetting; trained on the ROS-Gazebo simulation platform and further reduced the sim-to-real transfer gap through LiDAR clustering optimization.

CubeDN: Real-Time Drone Detection in 3D Space from Dual mmWave Radar Cubes

Yuan Fang, S. Julier

Object DetectionConvolutional Neural NetworkPoint Cloud

🎯 What it does: Proposes CubeDN, a single-stage end-to-end millimeter-wave radar detection network designed specifically for aerial drones, utilizing a dual radar configuration to achieve 3D detection.

Curb Your Attention: Causal Attention Gating for Robust Trajectory Prediction in Autonomous Driving

Ehsan Ahmadi, Amir Rasouli

Autonomous DrivingTransformer

🎯 What it does: Propose the CRiTIC model, which utilizes a causal discovery network to identify causal relationships between agents and employs a causal attention gate mechanism within a Transformer architecture to filter information, thereby enhancing the robustness of trajectory prediction.

Curiosity-Driven Imagination: Discovering Plan Operators and Learning Associated Policies for Open-World Adaptation

Pierrick Lorang, Matthias Scheutz

Robotic IntelligenceReinforcement LearningWorld Model

🎯 What it does: Proposed a hybrid planning and learning system that combines low-level neural networks learning stochastic transitions driven by Intrinsic Curiosity Module (ICM) for exploration, with high-level symbolic planning models capturing abstract transitions, enabling agents to plan and generate reward machines in an 'imagined' space, thus achieving rapid adaptation to open-world dynamic uncertain environments.

CurricuLLM: Automatic Task Curricula Design for Learning Complex Robot Skills Using Large Language Models

Kanghyun Ryu, Negar Mehr

Robotic IntelligenceTransformerLarge Language ModelReinforcement Learning

🎯 What it does: Proposed and implemented CurricuLLM, a system that leverages large language models to automatically generate task curricula for improving the learning efficiency of complex robotic skills.

Cycloidal Quasi-Direct Drive Actuator Designs with Learning-Based Torque Estimation for Legged Robotics

Alvin Zhu, Dennis W. Hong

Robotic IntelligenceReinforcement Learning

🎯 What it does: Designed and implemented a cycloidal quasi-direct drive actuator, and developed a torque estimation framework based on Actuator Network, aiming to enhance the performance of legged robots in high-torque and dynamic load tasks.

D3-ARM: High-Dynamic, Dexterous and Fully Decoupled Cable-Driven Robotic Arm

Hong Luo, Xueqian Wang

Robotic Intelligence

🎯 What it does: Designed and manufactured a fully decoupled, lightweight rope-driven robotic arm named D3-Arm, proposing a low-friction motion decoupling mechanism and rope pre-tensioning mechanism;

DA-Fusion: Deformable Attention-Based RGB-D Fusion Transformer for Unseen Object Instance Segmentation

Yesol Park, Byoung-Tak Zhang

SegmentationTransformerImageMultimodalityBenchmark

🎯 What it does: Proposed DA-Fusion, a deformable attention-based RGB-D fusion Transformer for unseen object instance segmentation, and released the Object Clutter Bin Dataset (OCBD) specifically designed for bin picking scenarios.

Da-Vil: Adaptive Dual-Arm Manipulation with Reinforcement Learning and Variable Impedance Control

Md Faizal Karim, Equal Contribution

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a pipeline that combines environment feedback-based policy learning with gradient optimization to learn control gains, enabling dual-arm adaptive manipulation.

DABI: Evaluation of Data Augmentation Methods Using Downsampling in Bilateral Control-Based Imitation Learning with Images

M. Kobayashi, Yuuki Uranishi

Data-Centric LearningRobotic IntelligenceImageMultimodalityTabular

🎯 What it does: Proposed and validated a bidirectional control-based imitation learning data augmentation method called DABI, achieving a tenfold increase in data volume through downsampling of high-frequency robot state data and low-frequency image data, and verified its effectiveness in real robot experiments.

DAP-LED: Learning Degradation-Aware Priors with Clip for Joint Low-Light Enhancement and Deblurring

Ling-Yun Wang, Lin Wang

RestorationTransformerVision Language ModelContrastive LearningImage

🎯 What it does: Propose a joint low-light enhancement and deblurring framework based on Transformer called DAP-LED, which utilizes CLIP to learn degradation levels of night images.

DARE: Diffusion Policy for Autonomous Robot Exploration

Yuhong Cao, G. Sartoretti

Robotic IntelligenceDiffusion model

🎯 What it does: Proposes the DARE scheme based on diffusion models, which can generate autonomous robot exploration paths with a single inference

Dark-DENet: A Lightweight Enhancement Network for Low-Light Object Detection

Xiaoyu Wu, Xinyu Jin

Object DetectionConvolutional Neural NetworkImage

🎯 What it does: Proposed a lightweight detection-driven enhancement network called Dark-DENet for target detection in low-light environments.

DART: Dexterous Augmented Reality Teleoperation Platform for Large-Scale Robot Data Collection in Simulation

Younghyo Park, Pulkit Agrawal

Robotic Intelligence

🎯 What it does: Proposes DART, a crowdsourcing teleoperation platform based on cloud simulation and AR, to redefine robotic data collection

Dashing for the Golden Snitch: Multi-Drone Time-Optimal Motion Planning with Multi-Agent Reinforcement Learning

Xian Wang, Shuo Li

OptimizationReinforcement Learning

🎯 What it does: Proposed a decentralized multi-agent reinforcement learning strategy network for time-optimal flight in multi-quadcopters, incorporating a soft collision avoidance mechanism.

Data Augmentation for NeRFs in the Low Data Limit

Ayush Gaggar, Todd D. Murphey

GenerationData SynthesisNeural Radiance Field

🎯 What it does: Propose a data augmentation method for NeRF under low data volume, adding training views by rejecting sampling from the posterior uncertainty distribution.

Data-Driven Sampling Based Stochastic MPC for Skid-Steer Mobile Robot Navigation

Ananya Trivedi, T. Padır

OptimizationRobotic Intelligence

🎯 What it does: Propose a data-driven sampling-based stochastic model predictive control method, utilizing a Gaussian process-enhanced dynamic single-track vehicle model to achieve sliding robot navigation; unify path tracking and obstacle avoidance through chance-constrained MPPI solving, and realize real-time performance with GPU acceleration.

Data-Efficient Learning from Human Interventions for Mobile Robots

Zhenghao Peng, Bolei Zhou

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Propose an online human-in-the-loop learning method called PVP4Real, combining imitation learning and reinforcement learning, which can efficiently learn mobile robot policies from real-time human interventions and demonstrations without requiring a reward function or pre-training.

Decentralized Drone Swaps for Online Rebalancing of Drone Delivery Tasks

K. Vakil, Alyssa Pierson

Optimization

🎯 What it does: A task allocation method based on expected demand is proposed for drone sharing between static warehouses.

Decentralized Nonlinear Model Predictive Control for Safe Collision Avoidance in Quadrotor Teams with Limited Detection Range

Manohari Goarin, Giuseppe Loianno

Autonomous DrivingOptimization

🎯 What it does: Proposed a decentralized nonlinear model predictive control (NMPC) integrated with exponential control barrier functions (ECBF) to achieve safe collision avoidance for multirotor drone teams in scenarios with limited detection range.

Decentralized Safe and Scalable Multi-Agent Control Under Limited Actuation

Vrushabh Zinage, E. Bakolas

OptimizationReinforcement Learning

🎯 What it does: Proposes an algorithm for achieving decentralized, safe, and scalable multi-agent control under limited execution capability.

Decentralized Vehicle Coordination: The Berkeley DeepDrive Drone Dataset and Consensus-Based Models

Fangyu Wu, A. Bayen

Autonomous DrivingExplainability and InterpretabilityImageVideoBenchmark

🎯 What it does: Collected and released a new dataset targeting road scenarios lacking explicit traffic rules, containing 20 aerial video clips, vehicle detection training images, trajectory estimation development tools, and proposed a consensus-based model for studying decentralized vehicle coordination and collision avoidance.

Decoupled Training Neural Solver for Dynamic Traveling Salesman Problem

Shaoheng Lin, Yancong Jia

OptimizationReinforcement Learning

🎯 What it does: Propose a decoupled training neural network solver (DTNS) for the Dynamic Traveling Salesman Problem (DTSP)

Deep Height Decoupling for Precise Vision-Based 3D Occupancy Prediction

Yuanwei Wu, Jian Yang

SegmentationAutonomous DrivingSupervised Fine-TuningImageBenchmark

🎯 What it does: Proposed the Deep Height Decoupling (DHD) framework, which predicts height maps and uses Mask Guided Height Sampling (MGHS) to separate image features by height ranges, followed by a Synergistic Feature Aggregation (SFA) module to enhance feature representation, thereby achieving more accurate vision-based 3D occupancy prediction.

Deep Learning Based Topography Aware Gas Source Localization with Mobile Robot

Changhao Tian, Xiaodong Chen

Robotic IntelligenceConvolutional Neural NetworkSimultaneous Localization and MappingMultimodality

🎯 What it does: Propose a gas source localization method based on deep learning, integrating SLAM with U-Net, utilizing gas sensors, wind speed, and 2D occupancy maps to predict the gas source location.

Deep Learning-Based Friction Compensation in Low Velocity for Enhanced Direct Teaching in Collaborative Manipulators

Seohyun Choi, Wan Kyun Chung

Robotic Intelligence

🎯 What it does: Propose a low-speed friction compensation framework based on deep neural networks to enhance the direct teaching method of collaborative robotic arms, improving friction estimation and training effectiveness through preprocessing algorithms and a custom loss function.

Deep Learning-Enhanced Visual Monitoring in Hazardous Underwater Environments with a Swarm of Micro-Robots

Shuang Chen, Amir Atapour-Abarghouei

RestorationData SynthesisPose EstimationRobotic IntelligenceImage

🎯 What it does: Propose a method that combines data simulation and a multi-modal deep learning network for coordinate prediction and image reconstruction to address the visual monitoring problem of underwater micro-robot swarms under drift and rotation disturbances.

Deep Reinforcement Learning for Coordinated Payload Transport in Biped-Wheeled Robots

Dhruv K Mehta, V. Krovi

Robotic IntelligenceReinforcement Learning

🎯 What it does: Studied the collaborative transportation of a load by two bipedal wheeled robots, generating robot motion planning through deep reinforcement learning.

Deep Reinforcement Learning-Based Semi-Autonomous Control for Magnetic Micro-Robot Navigation with Immersive Manipulation

Yudong Mao, Dandan Zhang

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a semi-automated control framework based on deep reinforcement learning (DRL-SC) for navigation of magnetic microrobots in a simulated microvascular system, integrated with mixed reality (MR) to enable immersive control;

DeepVL: Dynamics and Inertial Measurements-based Deep Velocity Learning for Underwater Odometry

Mohit Singh, Kostas Alexis

Robotic IntelligenceRecurrent Neural NetworkSimultaneous Localization and MappingTime SeriesSequential

🎯 What it does: Propose a dynamic perception-based learning model that predicts the underwater robot's self-speed and its uncertainty by combining recurrent neural networks with inertial measurements, motor commands, battery voltage, and the hidden state from the previous time step. The model integrates network ensembles with an extended Kalman filter (EKF) to fuse inertial and barometer information, achieving long-term underwater odometry integration with visual-inertial odometry.

Deformable Gaussian Splatting for Efficient and High-Fidelity Reconstruction of Surgical Scenes

Jiwei Shan, Hesheng Wang

Gaussian Splatting

🎯 What it does: Developed the EH-SurGS algorithm, achieving efficient and high-fidelity reconstruction of deformable surgical scenes using 3D Gaussian light scattering; the algorithm models the lifecycle of 3D Gaussians to capture both regular and irreversible deformations, and employs an adaptive motion hierarchy strategy to distinguish static and deformable regions, reducing the number of Gaussians passing through deformation fields to enhance rendering speed.

Deformable Multibody Modeling for Model Predictive Control in Legged Locomotion with Embodied Compliance

Keran Ye, Konstantinos Karydis

OptimizationRobotic Intelligence

🎯 What it does: A method for dynamic gait stabilization in compliant humanoid robots with ontological compliance is proposed, integrating a unified description of rigid and flexible objects and modeling deformable multibody systems, while developing a Composite Centroidal Predictive Deformation Inertia (CCPDI) tensor to be integrated into a standard MPC framework.

Deformpam: Data-Efficient Learning for Long-Horizon Deformable Object Manipulation Via Preference-Based Action Alignment

Wendi Chen, Cewu Lu

Robotic IntelligenceReinforcement Learning from Human FeedbackDiffusion modelPoint Cloud

🎯 What it does: Propose the DeformPAM framework, which decomposes long-term deformable object manipulation tasks by leveraging preference learning and reward-driven action selection. It uses 3D point cloud inputs and diffusion models to model action distributions, while training an implicit reward model to select optimal actions.

Delayed-Decision Motion Planning in the Presence of Multiple Predictions

David Isele, Sangjae Bae

Autonomous DrivingOptimization

🎯 What it does: Propose a behavior planning scheme under multiple predicted future scenarios and present its maximum entropy formalization.

Deliberative Control-Aware Motion Planning for Kinematic-Constrained UAVs in a Dynamic Environment

Elias J. R. Freitas, Luciano C. A. Pimenta

Autonomous DrivingOptimization

🎯 What it does: Proposes a motion planning method for unmanned vehicles in dynamic environments that uses NURBS curves to represent paths, employs differential evolution algorithms to optimize curve parameters and speed, and ensures collision safety through velocity obstacle constraints.

DELTA: Decomposed Efficient Long-Term Robot Task Planning Using Large Language Models

Yuchen Liu, Marco Aiello

Robotic IntelligenceLarge Language ModelGraph

🎯 What it does: Proposed and implemented a robot task planning framework called DELTA based on large language models (LLMs), which utilizes scene graphs to rapidly generate precise planning problem descriptions within LLMs and decomposes long-term task goals into autoregressive subgoals, significantly enhancing the feasibility and efficiency of planning.

DemoStart: Demonstration-Led Auto-Curriculum Applied to Sim-to-Real with Multi-Fingered Robots

Maria Bauzá, N. Heess

Domain AdaptationRobotic IntelligenceReinforcement LearningMultimodality

🎯 What it does: Developed DemoStart, an automatic curriculum reinforcement learning framework that learns complex control behaviors on a three-finger robotic arm using sparse rewards and a small number of simulated demonstrations, achieving zero-shot sim-to-real transfer.

Dense Fixed-Wing Swarming Using Receding-Horizon NMPC

V. Madabushi, Joseph L. Moore

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: A control method for close-proximity cooperative flight of fixed-wing UAVs based on recursive time-domain nonlinear model predictive control (NMPC) is proposed, with the introduction of a statistical upper bound of collision probability and a new evaluation metric, verified through simulation and hardware experiments.

DENSER: 3D Gaussian Splatting for Scene Reconstruction of Dynamic Urban Environments

Mahmud A. Mohamad, Raphaël Frank

Autonomous DrivingGaussian SplattingPoint Cloud

🎯 What it does: Proposes the DENSER framework based on 3D Gaussian splatting for dynamic urban environment reconstruction, significantly enhancing the appearance and shape modeling of foreground dynamic objects.

Deploying Ten Thousand Robots: Scalable Imitation Learning for Lifelong Multi-Agent Path Finding

He Jiang, Jiaoyang Li

OptimizationRobotic Intelligence

🎯 What it does: Proposed a multi-agent path planning solver based on imitation learning called SILLM, integrating a novel communication module, systematic single-step collision resolution, and global guidance techniques;

Depth Estimation Based on 3D Gaussian Splatting Siamese Defocus

Jinchang Zhang, Guoyu Lu

Depth EstimationGaussian SplattingImage

🎯 What it does: A self-supervised framework based on 3D Gaussian splatting and a Siamese network is proposed for monocular depth estimation. The framework learns the blur degree of the same scene at different focal lengths within a focal stack, enabling the prediction of a defocus map (Defocus Map) and circular blur radius (CoC) from a single blurry image. The defocus map is then used as input to DepthNet for depth inference. Subsequently, the 3D Gaussian splatting model renders the blurry image based on the predicted CoC, and the rendered result is compared with the real blurry image to provide additional self-supervised signals to the Siamese network.

Depth Estimation Through Translucent Surfaces

Siyu Dai, Sisir Karumanchi

RestorationDepth EstimationRobotic IntelligenceImage

🎯 What it does: The study proposes two methods for depth estimation through translucent barriers: first, using image inpainting to remove the translucent barrier followed by standard monocular or stereo depth estimation models; or directly training depth models on images containing the translucent barrier.

Depth Restoration of Hand-Held Transparent Objects for Human-to-Robot Handover

Ran Yu, Wenbo Ding

RestorationPose EstimationDepth EstimationRobotic IntelligenceNeural Radiance FieldImage

🎯 What it does: Proposes a gesture-aware depth recovery method HADR based on a single RGB-D image, utilizing hand pose to guide implicit neural representations for reconstructing the depth of transparent handheld objects

Depth-Temporal Attention with Dual Modality Data for Walking Intention Prediction in Close-Proximity Front-Following

Chongyu Zhao, Wu Chuan

Domain AdaptationRobotic IntelligenceTransformerMultimodality

🎯 What it does: Proposes a scheme based on a deep-temporal attention network, utilizing lower-limb depth images and robot motion signals to predict human walking intent, thereby achieving close-range front-following.

Design and Effectiveness of Virtual Monitors and AR-Based Endoscope Control for Robotically Assisted Laparoscopic Surgery

Nikola Budjakoski, Julian Klodmann

Robotic IntelligenceBiomedical Data

🎯 What it does: Integrate robotic endoscope platform with augmented reality display system to enable direct control of the endoscope and provide an interactive 3D virtual monitor.

Design and Evaluation of High-Performance Motion-Decoupled Cable Transmission Modules

Ryo Takei, J. P. Whitney

Robotic IntelligenceMagnetic Resonance Imaging

🎯 What it does: Proposed a motion decoupling cable transmission module and demonstrated its applications in robotic arms and MRI-safe needle biopsy robots.

Design and Experimental Validation of Woodwork-Inspired Soft Pneumatic Grippers

Abriana Stewart-Height, H. Asada

Robotic Intelligence

🎯 What it does: Designed and verified a pair of soft grippers with an inverted mortise and tenon structure, achieving the ability to carry heavy loads using low-pressure gas through soft finger interlocking

Design and Implementation of a Snake Robot for Cranial Surgery

Jones Law, D. Podolsky

Robotic Intelligence

🎯 What it does: Designed and implemented a dual-segment flexible snake robot for minimally invasive craniectomy, equipped with a bone hammer, dura/skin retractor, and channels for endoscopes and surgical instruments.

Design and Implementation of a Swimming and Walking Quadruped for Seafloor Exploration

A. Chase, Jonathan E. Clark

Robotic Intelligence

🎯 What it does: Designed and implemented a quadruped robot capable of walking on the seabed and swimming in intermediate fluid, demonstrating its multimodal trajectory execution.

Design of a Bioinspired Jumping Mechanism for Self-Takeoff of Flapping Robot

Erzhen Pan, W. Xu

Robotic Intelligence

🎯 What it does: Designed and tested a bio-inspired lightweight single-degree-of-freedom jumping leg to enable self takeoff for a flapping-wing robot.

Design of a Formation Control System to Assist Human Operators in Flying a Swarm of Robotic Blimps

Tianfu Wu, Fumin Zhang

Robotic IntelligenceImage

🎯 What it does: Proposed and experimentally verified a dynamic leader-switching pioneer-follower formation control system for indoor airships

Design of a Novel Pneumatic Soft Gripper for Robust Adaptive Grasping

Xiantao Sun, Weihai Chen

Robotic Intelligence

🎯 What it does: A novel pneumatic three-finger soft gripper is proposed to achieve robust adaptive grasping

Design of an Articulated Modular Caterpillar Using Spherical Linkages

Sam O'Connor, Mark M. Plecnik

OptimizationRobotic Intelligence

🎯 What it does: Designed a modular caterpillar robot using spherical links, and achieved segment movement by selecting a single-degree-of-freedom curve in SO(3).

Design, Contact Modeling, and Collision-Inclusive Planning of a Dual-Stiffness Aerial RoboT (DART)

Yogesh Kumar, Wenlong Zhang

Robotic Intelligence

🎯 What it does: Designed and tested a dual-stiffness quadrotor robot (DART) that can switch between rigid and flexible modes after collision through a locking mechanism, and proposed a collision response prediction model based on linear complementarity system theory.

Design, Implementation, and Validation of an Ungrounded Visuo-Tactile Haptic Interface for Robotic Teleoperation in High-Risk Steel Production

Jaehyun Park, Keehoon Kim

Robotic IntelligenceVision-Language-Action ModelMultimodality

🎯 What it does: Proposed and implemented a non-grounded visuo-haptic interface called POstick-VF, designed for high-risk steel production tasks.

Designing Experimental Setup Emulating Log-Loader Manipulator and Implementing Anti-Sway Trajectory Planner

Iman Jebellat, Inna Sharf

OptimizationRobotic Intelligence

🎯 What it does: Designed an experimental platform to simulate the motion of a wood loader's grab, using a Kinova Jaco2 robotic arm and a custom-made end-effector, and implemented anti-swing trajectory planning via dynamic programming, verifying performance under different point-to-point motions through multiple repeated experiments.

DetailRefine: Towards Fine-Grained and Efficient Online Monocular 3D Reconstruction

Fupeng Chu, Ronghan Chen

Depth EstimationComputational EfficiencyImage

🎯 What it does: Propose dynamic detail refinement (DDR), discriminability-aware fusion (DAF), and hierarchical hybrid sparsification (HHS) methods to enhance the fine-grained detail and efficiency of online monocular 3D reconstruction.

Detecting Perception-Based Attacks using Visual Odometry: Inconsistency Modeling and Checking on Robotic States

Yuan Xu, Tianwei Zhang

Anomaly DetectionRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed a method to detect perception attacks by modeling the inconsistency between the robot's physical state and estimated state.

Detection of Fast-Moving Objects with Neuromorphic Hardware

Andreas Ziegler, Andreas Zell

Object DetectionRobotic IntelligenceSpiking Neural NetworkBenchmark

🎯 What it does: Provides an overview of real-time processing of Spiking Neural Networks (SNN) on neuromorphic hardware, benchmarks event-driven object detection on three popular devices, and implements real-time SNN operation in a ping-pong robot environment.

Development of a New Biped Robot with Adaptive Suction Modules for Climbing on Curved Surfaces

Zikang Li, Qingsong Xu

Robotic Intelligence

🎯 What it does: A bipedal curved surface climbing robot (BCCR) with five-degree-of-freedom motion is proposed, equipped with adaptive vacuum suction modules capable of adhering to and smoothly climbing both curved surfaces and planes.

Development of Contactless Delivery Service Robot with Modular Working Platform in Isolation Wards

Kyon-Mo Yang, Kapho Seo

Robotic Intelligence

🎯 What it does: Proposed a modular robot system integrating a working platform and a mobile base for preventing cross-infection in isolation wards, as well as enabling item delivery and waste transportation;

Development of Multi-Joint Biohybrid Soft Robot by Using Skeletal Muscle Tissue

Eunhye Kim, T. Fukuda

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed modular bioactuators using skeletal muscle tissue, and utilized them to construct multi-jointed soft robots, achieving for the first time a multi-degree-of-freedom rotational joint manipulator.

DexMimicGen: Automated Data Generation for Bimanual Dexterous Manipulation via Imitation Learning

Zhenyu Jiang, Yuke Zhu

Robotic IntelligenceSequential

🎯 What it does: Developed the DexMimicGen system, which automatically generates training trajectories for dual-arm dexterous robots using a small number of human demonstrations

Dexterous Assembly Using a Planar Hand Having Programmable Passive Compliance

Jacob M. Frye, J. Schimmels

Robotic Intelligence

🎯 What it does: Proposed and verified a programmable three-finger antagonistic compliant hand (P3ACH), capable of achieving predetermined compliance behaviors in a multidirectional compliance space, and demonstrated its operational flexibility through various assembly tasks.

Dexterous Three-Finger Gripper based on Offset Trimmed Helicoids (OTHs)

Qinghua Guan, Josie Hughes

Robotic Intelligence

🎯 What it does: Designed and tested a three-fingered gripper based on offset-trimmed helix (OTH) to mimic the flexibility of human fingers

Dexterous Ungrasping Manipulation in Three Dimensions

Taewoong Kang, Jungwon Seo

Robotic Intelligence

🎯 What it does: Developed a 3D non-grasping (placement) technique that achieves object rolling and/or sliding placement through non-static contact, applicable to slender objects;

DFM: Deep Fourier Mimic for Expressive Dance Motion Learning

Ryo Watanabe, Marco Hutter

GenerationReinforcement Learning

🎯 What it does: Proposes a dance motion learning method based on deep Fourier imitation, combining reinforcement learning to achieve smooth action transitions while supporting auxiliary tasks such as walking and gaze.

DGS-SLAM: A Visual Dense SLAM Based on Gaussian Splatting in Dynamic Environments

Yushi Chen, Haiyong Luo

Gaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Developed a visual dense SLAM system called DGS-SLAM based on Gaussian scattering, which can achieve robust localization and high-fidelity static map reconstruction in dynamic environments.

DGTR: Distributed Gaussian Turbo-Reconstruction for Sparse-View Vast Scenes

Hao Li, Junwei Han

RestorationGenerationComputational EfficiencyGaussian SplattingImage

🎯 What it does: Proposed a distributed and efficient Gaussian reconstruction framework called DGTR for large-scale scene reconstruction and novel view synthesis from sparse perspectives.

Diff-Dagger: Uncertainty Estimation With Diffusion Policy for Robotic Manipulation

Sung-Wook Lee, Yen-Ling Kuo

Robotic IntelligenceDiffusion model

🎯 What it does: Proposed a robot-gated DAgger algorithm called Diff-DAgger, combining the training objectives of diffusion policies with interactive imitation learning to enhance uncertainty estimation and performance in robot operations;

Diff-Refiner: Enhancing Multi-Agent Trajectory Prediction with a Plug-and-Play Diffusion Refiner

Xiangzheng Zhou, Jian Yang

Autonomous DrivingDiffusion modelOrdinary Differential Equation

🎯 What it does: This paper proposes a Diffusion Refiner, which first uses a baseline model to predict rough trajectories and then refines them using a diffusion model, thereby improving the performance of multi-agent trajectory prediction.

DiffCP: Ultra-Low Bit Collaborative Perception via Diffusion Model

Ruiqing Mao, Zhisheng Niu

CompressionAutonomous DrivingDiffusion model

🎯 What it does: Propose a collaborative perception framework DiffCP based on diffusion models, which can achieve feature-level collaboration with extremely low communication costs

Differentiable Composite Neural Signed Distance Fields for Robot Navigation in Dynamic Indoor Environments

S. Bukhari, A. H. Qureshi

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Proposed a differentiable composite neural signed distance field (SDF) framework that enables robot navigation in dynamic indoor environments using onboard RGB-D sensors.

Diffusion Based Robust LiDAR Place Recognition

Benjamin Krummenacher, Marco Hutter

Pose EstimationRetrievalGraph Neural NetworkDiffusion modelPoint Cloud

🎯 What it does: Achieving global localization and relocalization for robots in construction sites using LiDAR data and diffusion models

Diffusion Meets Options: Hierarchical Generative Skill Composition for Temporally-Extended Tasks

Zeyu Feng, Harold Soh

GenerationReinforcement LearningDiffusion model

🎯 What it does: Proposed the Doppler framework to achieve hierarchical plan generation and updating based on linear temporal logic (LTL) instructions, and utilized diffusion models to generate low-level actions as well as deterministic guided posterior sampling techniques to enhance generation speed and diversity.

Diffusion-Based Generative Models for 3D Occupancy Prediction in Autonomous Driving

Yunsheng Wang, Hang Zhao

GenerationAutonomous DrivingDiffusion modelImage

🎯 What it does: Using visual input to predict 3D occupancy grids

Diffusion-Based Self-Supervised Imitation Learning from Imperfect Visual Servoing Demonstrations for Robotic Glass Installation

Canran Xiao, Wenrui Chen

Robotic IntelligenceDiffusion model

🎯 What it does: Using diffusion models for self-supervised imitation learning, based on imperfect visual servo demonstrations to achieve glass installation.

Diffusion-Informed Probabilistic Contact Search for Multi-Finger Manipulation

Abhinav Kumar, Dmitry Berenson

OptimizationRobotic IntelligenceDiffusion model

🎯 What it does: Proposed a probability contact search method called DIPS based on diffusion models, which plans contact pattern sequences for multi-finger grasping through A* search, and improves planning quality by combining particle filter-inspired mutation estimation and discriminators.

Digiforests: a Longitudinal Lidar Dataset for Forestry Robotics

Meher V. R. Malladi, Maurice F. Fallon

Robotic IntelligencePoint CloudBenchmarkAgriculture Related

🎯 What it does: Provides a real-world long-term LiDAR forest dataset containing multiple records of the same forest area, along with semantic, instance, and fine-grained annotations for tree trunks and crowns, and reference field measurements.

Digital Beamforming Enhanced Radar Odometry

Jingqi Jiang, Sen Wang

Pose EstimationSimultaneous Localization and Mapping

🎯 What it does: Developed a radar signal processing pipeline integrating spatial domain beamforming technology and extended it to 3D angle estimation;

Digital Model-Driven Genetic Algorithm for Optimizing Layout and Task Allocation in Human-Robot Collaborative Assemblies

Christian Cella, Paolo Rocco

OptimizationRobotic Intelligence

🎯 What it does: Propose an iterative optimization scheme that uses digital models and genetic algorithms to jointly optimize layout and task allocation in human-robot collaborative assembly workcells.

Digital Twins Meet the Koopman Operator: Data-Driven Learning for Robust Autonomy

Chinmay Samak, V. Krovi

Data SynthesisAutonomous DrivingOptimizationWorld Model

🎯 What it does: Generate precise digital models of vehicles and their target operating conditions using digital twin technology, construct an off-road vehicle dynamics model based on simulation data using Koopman operator theory, and apply this model to local motion planning and optimal vehicle control.

Directed-CP: Directed Collaborative Perception for Connected and Autonomous Vehicles via Proactive Attention

Yihang Tao, Yuguang Fang

Autonomous DrivingPoint Cloud

🎯 What it does: Proposes a directed active directional perception collaborative perception system called Directed-CP, which allows the ego vehicle to actively specify directions of interest and reallocate attention to enhance local directional perception performance.

Direction Informed Trees (DIT*): Optimal Path Planning via Direction Filter and Direction Cost Heuristic

Liding Zhang, A. Knoll

OptimizationRobotic Intelligence

🎯 What it does: Proposed the Directional Information Tree (DIT*) sampling planner, achieving goal-biased exploration by optimizing search directions for each edge and introducing directional filters and directional cost heuristics.

DISC: Dataset for Analyzing Driving Styles in Simulated Crashes for Mixed Autonomy

Sandip Sharan Senthil Kumar, Ming C. Lin

Data SynthesisAutonomous DrivingTime Series

🎯 What it does: Constructed the DISC dataset, recording the driving styles of hundreds of drivers facing 12 simulated accident scenarios in a virtual city.

Discovering Object Attributes by Prompting Large Language Models With Perception-Action Apis

A. Mavrogiannis, Y. Aloimonos

RecognitionRobotic IntelligenceLarge Language ModelVision Language ModelVision-Language-Action ModelImageMultimodality

🎯 What it does: Construct a perception-action API using large language models (LLM) and vision-language models (VLM), generating programs via LLM to actively perceive and identify non-visual attributes in images.

Discrete Contrastive Learning for Diffusion Policies in Autonomous Driving

Kalle Kujanpää, Ville Kyrki

Autonomous DrivingDiffusion modelContrastive Learning

🎯 What it does: Propose a method that utilizes contrastive learning to extract a driving style dictionary, discretizes the style through quantization, and then employs a conditional diffusion strategy to simulate human driver behavior.

Discrete Policy: Learning Disentangled Action Space for Multi-Task Robotic Manipulation

Kun Wu, Jian Tang

Robotic IntelligenceVision-Language-Action ModelImageTextMultimodality

🎯 What it does: Studies visual motion policies for multi-task robotic manipulation, proposes the Discrete Policy method, which maps action sequences into a discrete latent space using vector quantization, learns task-specific codes, and reconstructs them into actions.

Distributed Certifiably Correct Range-Aided SLAM

Alexander Thoms, Sriram Narasimhan

OptimizationSimultaneous Localization and Mapping

🎯 What it does: Proposed a distributed, trustworthy globally optimal RA-SLAM algorithm called DCORA

Distributed Invariant Kalman Filter for Object-Level Multi-Robot Pose SLAM

Haoying Li, Junfeng Wu

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Propose a distributed invariant Kalman filter based on covariance intersection (CI) for multi-robot pose estimation, and adopt an object-level measurement model to reduce communication burden.

Distributed Loitering Synchronization with Fixed-Wing UAVs

Ahmed AlKatheeri, E. Ferrante

Robotic Intelligence

🎯 What it does: Study distributed standby synchronization algorithms, propose two new methods, MOSA and FPS, and evaluate their synchronization performance against baseline distributed consensus algorithms on circular orbits of fixed-wing drones; validate their effectiveness through simple simulations, Gazebo simulations (considering fixed-wing dynamics), and real flight experiments with three fixed-wing drones.

Distributed Multi-Robot Source Seeking in Unknown Environments with Unknown Number of Sources

Lingpeng Chen, Woojun Kim

Robotic Intelligence

🎯 What it does: Proposes a distributed multi-robot source search framework named DIAS, applicable to environments where the number of sources is unknown and may exceed the number of robots.

Distributed Perception Aware Safe Leader Follower System via Control Barrier Methods

Richie R. Suganda, Bin Hu

Autonomous DrivingOptimizationConvolutional Neural NetworkImage

🎯 What it does: A distributed leader-follower safety control scheme considering camera field-of-view (FOV) limitations is proposed, utilizing control barrier functions (CBF) to ensure forward invariance of the safe set.

Distributed Pursuit of an Evader with Adaptive Robust Path Control Under State Measurement Uncertainty

Kai Rao, Yunkai Lv

Optimization

🎯 What it does: Proposes a distributed tracking framework that considers the uncertainty of state measurements, mainly consisting of two parts: safety tracking area calculation based on Voronoi cells and an adaptive robust path controller based on control barrier functions.

DiTer++: Diverse Terrain and Multi-Modal Dataset for Multi-Robot SLAM in Multi-Session Environments

Juwon Kim, Younggun Cho

Simultaneous Localization and MappingMultimodalityBenchmark

🎯 What it does: Proposed the DiTer++ dataset for large-scale SLAM research in multi-robot, multi-session environments.

Diver to Robot Communication Underwater

Robert Codd-Downey, Michael R. M. Jenkin

RecognitionRobotic IntelligenceRecurrent Neural Network

🎯 What it does: Propose an interactive process for unmanned underwater vehicles to identify and confirm standard SCUBA gestures and their sequences, enabling gesture communication between divers and robots.

Diverse Motion Planning with Stein Diffusion Trajectory Inference

Zeya Yin, Fabio Ramos

OptimizationRobotic IntelligenceDiffusion model

🎯 What it does: Integrate Stein Variational Gradient Descent (SVGD) with Gaussian Process Motion Planning (GPMP), utilizing diffusion models as a multimodal prior and incorporating path signatures to enhance the diversity of the posterior distribution, aiming to rapidly approximate the trajectory posterior distribution;

Do Looks Matter? Exploring Functional and Aesthetic Design Preferences for a Robotic Guide Dog

Aviv L. Cohav, Bruce N. Walker

Robotic IntelligenceText

🎯 What it does: Data was collected through interviews and questionnaires to identify and propose functional and aesthetic design concepts for robotic guide dogs, determining essential and preferred characteristics for future robotic guide dogs.

Does Bilevel Optimization Result in More Competitive Racing Behavior?

Andrew Cinar, Forrest Laine

OptimizationPhysics Related

🎯 What it does: Study the competitive dynamic game of two-vehicle racing, construct a new two-player vehicle racing model, and conduct large-scale empirical studies to compare the competitive performance of different solutions.