arXivSub Start free trial

ICRA 2025 Papers — Page 6

IEEE International Conference on Robotics and Automation · 1604 papers

Equivariant Filter for Tightly Coupled LiDAR-Inertial Odometry

Anbo Tao, Xingxing Li

Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a tightly coupled LiDAR-IMU pose estimation method called Eq-LIO based on the equivariant filter (EqF), which can achieve naturally consistent and highly robust state estimation under nonlinear models.

ERetinex: Event Camera Meets Retinex Theory for Low-Light Image Enhancement

Xuejian Guo, Yue Gao

RestorationImage

🎯 What it does: Propose a low-light image enhancement method called ERetinex by combining event cameras with Retinex theory

Ergodic Exploration over Meshable Surfaces

Dayi Dong, Ian Abraham

OptimizationMesh

🎯 What it does: Extend ergodic search to surfaces approximated by arbitrary triangular meshes;

Ergodic Trajectory Optimization on Generalized Domains Using Maximum Mean Discrepancy

Christian Hughes, Ian Abraham

Optimization

🎯 What it does: Proposed a general-domain Euclidean trajectory optimization method using kernel maximum mean discrepancy (MMD) for generating coverage paths

Error-Subspace Transform Kalman Filter Based Real-Time Gait Prediction for Rehabilitation Exoskeletons

Haozhou Zeng, Tao Liu

Pose EstimationRobotic IntelligenceTime SeriesSequentialBiomedical Data

🎯 What it does: Proposes a real-time gait prediction method based on Bayesian inference, utilizing von Mises basis functions to represent periodic gait information, and employing an error subspace transform Kalman filter (ESTKF) to predict gait trajectories, followed by a fully connected neural network (FCNN) for real-time estimation of walking speed.

Escaping Local Minima: Hybrid Artificial Potential Field with Wall-Follower for Decentralized Multi-Robot Navigation

Joonkyung Kim, Changjoo Nam

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a hybrid method combining wall-following behavior with artificial potential fields (APF) to achieve mapless, decentralized multi-robot navigation using only local sensor information and avoiding local minima.

Estimating Commonsense Scene Composition on Belief Scene Graphs

M. A. Saucedo, G. Nikolakopoulos

Representation LearningGraph Neural NetworkLarge Language ModelGraph

🎯 What it does: Proposed and implemented the concept of common-sense scene composition, extended the belief scene graph to estimate the spatial distribution of unseen objects, and modeled spatial relationships between related objects through a joint probability distribution.

Estimating Control Barriers from Offline Data

Hong-Den Yu, Sicun Gao

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Propose a new framework that learns neural control barrier functions (CBFs) using sparsely labeled offline data, expands labels through discrete distribution analysis, eliminates reliance on high-performance expert controllers, enables data collection with suboptimal or manual control, and implements dynamic obstacle avoidance on real platforms;

Estimating High-Resolution Neural Stiffness Fields Using Visuotactile Sensors

Jiaheng Han, Kris Hauser

Neural Radiance FieldMultimodality

🎯 What it does: Utilizing the Punyo bubble visual tactile sensor, a dense contact force observation model is constructed, and a neural volume stiffness field (VSF) is proposed to achieve high-resolution stiffness estimation.

Estimating User Engagement in Human Robot Interaction Using a Dynamic Bayesian Network

Xiaoxuan Hei, Adriana Tapus

ClassificationMultimodality

🎯 What it does: Using dynamic Bayesian networks (DBN) combined with multimodal features from visual and thermal imaging (head rotation, eye movement, facial expressions, and facial temperature) to estimate user engagement in human-computer interaction.

ETSM: Automating Dissection Trajectory Suggestion and Confidence Map-Based Safety Margin Prediction for Robot-Assisted Endoscopic Submucosal Dissection

Mengya Xu, Hongliang Ren

Robotic IntelligenceVideoBiomedical Data

🎯 What it does: Constructed the ETSM dataset and proposed RCMNet, which automatically suggests resection trajectories and predicts safety margins based on confidence maps.

Evaluating Global Geo-Alignment for Precision Learned Autonomous Vehicle Localization Using Aerial Data

Yi Yang, Will Maddern

Autonomous DrivingSimultaneous Localization and MappingImage

🎯 What it does: Studied the impact of improving the alignment between aerial maps and vehicle sensor data during training on the accuracy of learned localization, compared two alignment methods, and evaluated the learning-based localization system on a 1600 km autonomous driving dataset.

Evaluating Human-Robot Skill Gaps in Electrical Circuit Inspection: A New Electronic Task Board for Benchmarking Manipulation

Peter So, Sami Haddadin

Robotic IntelligenceBenchmark

🎯 What it does: Introduces a real-world benchmark based on internet-connected electronic task boards for evaluating a robot's operational skills in circuit detection, collects contributions from 16 teams across 6 tasks and the best performance of 30 human subjects; conducts timing studies and provides example robot solutions.

Evaluating Robotic Performative Autonomy in Collaborative Contexts Impacted by Latency

R. S. Silva, Tom Williams

Robotic Intelligence

🎯 What it does: Assessed the impact of communication delay and executional autonomy (PA) on team trust, situation awareness (SA), and humans' perception of robot intelligence and autonomy.

eViper-2D: A Thin Large-Area Soft Robotics Platform

Hsin Cheng, Minjie Chen

OptimizationRobotic Intelligence

🎯 What it does: This paper presents the key principles of the eViper-2D soft robot platform, elaborates its mechanical, electrical, and control frameworks, and develops a systematic and scalable method to study the impact of different driving modes on robot motion dynamics and energy efficiency. By integrating power electronics, communication circuits, piezoelectric actuators, and batteries, eViper-2D achieves rapid design iteration and quick evaluation of control strategies for multi-actuator soft robots. The platform also supports automated data acquisition for data-driven modeling, enabling optimization of driving modes to achieve agile motion and minimal transportation costs.

EVLoc: Event-Based Visual Localization in LiDAR Maps via Event-Depth Registration

Kuangyi Chen, Friedrich Fraundorfer

Pose EstimationConvolutional Neural NetworkOptical FlowPoint Cloud

🎯 What it does: This paper proposes a visual localization method based on an event camera, which utilizes existing LiDAR maps for pose refinement. The process includes first projecting LiDAR point clouds into 2D to obtain a depth map with a rough initial pose, then aligning events with the 2D depth map using an optical flow estimation network, followed by estimating the camera pose via a PnP solver. Additionally, the authors introduce a novel frame-based event representation to enhance structural clarity and design an auxiliary variable prediction module as a regularization term to mitigate the impact of biases in real poses on network convergence.

EvMAPPER: High-Altitude Orthomapping with Event Cameras

Fernando Cladera, Vijay Kumar

GenerationImage

🎯 What it does: Proposed the first method for generating high-altitude orthomosaics based on event cameras

Exact Imposition of Safety Boundary Conditions in Neural Reachable Tubes

Aditya Singh, Somil Bansal

OptimizationReinforcement Learning

🎯 What it does: Propose ExactBC, a variant of DeepReach, which precisely enforces safety boundary conditions during the learning process by reconstructing the overall value function as a weighted sum of boundary conditions and NN outputs, eliminating the dependence on boundary loss terms.

Excavating in the Wild: The GOOSE-Ex Dataset for Semantic Segmentation

Raphael Hagmanns, J. Petereit

SegmentationRobotic IntelligenceVideoMultimodalityBenchmark

🎯 What it does: Developed and made public the GOOSE-Ex dataset, containing 5,000 frames of multimodal annotated data from robotic excavators and quadrupedal platforms, and conducted a comprehensive analysis of semantic segmentation performance across different platforms and sensor modalities in unseen environments.

Expert-Enhanced Masked Point Modeling for Point Cloud Self-Supervised Learning

YuJun Liu, Shu-Tao Xia

Representation LearningMixture of ExpertsPoint Cloud

🎯 What it does: Introduce an expert-enhanced masked point modeling method in self-supervised learning for point clouds, achieving routing and analysis of different semantics by inserting a Sparse Mixture of Experts (SMoE) layer after each backbone block.

Explainable Reinforcement Learning via Dynamic Mixture Policies

Maximilian Schier, Bodo Rosenhahn

Explainability and InterpretabilityReinforcement Learning

🎯 What it does: Proposed an interpretable family of control strategies, constructing readable stochastic policies by utilizing diagonal compressed Gaussian distributions and categorical mixture distributions on decomposed observation subspaces;

Exploration and Analysis of Torso-Limb Coordination of Quadruped Walkers with Compliant Torso

Yuxuan Xiang, Fumihiko Asano

Robotic Intelligence

🎯 What it does: The study investigates the walking performance and trunk-limb coordination mechanisms of a flexible-trunk quadruped robot under front limb or hind limb driving through numerical simulation.

Explore the LiDAR-Camera Dynamic Adjustment Fusion for 3D Object Detection

Yiran Yang, Jingdong Wang

Object DetectionAutonomous DrivingImageMultimodalityPoint Cloud

🎯 What it does: Proposes a LiDAR-Camera dynamic adjustment fusion technique, enhancing 3D object detection through a three-phase domain alignment module, modality interaction with specialized enhancement, and semantic-geometric adaptive learning.

Exploring Adversarial Obstacle Attacks in Search-Based Path Planning for Autonomous Mobile Robots

Adrian Szvoren, Nilufer Tuptuk

OptimizationAdversarial AttackRobotic Intelligence

🎯 What it does: This paper investigates the robustness of the A* search algorithm when facing adversarial obstacle attacks, using malicious software to conduct simulations on TurtleBot in Gazebo and performing real-world deployment on the Unitree Go1 robot to evaluate the impact of attacks on path planning.

Exploring the Domain-Invariant Flow Representation in Vision-Based Tactile Sensors for Omni-Hardness Perception

Xuewen Yang, Aiguo Song

Domain AdaptationRepresentation LearningRobotic IntelligenceFlow-based ModelImage

🎯 What it does: Proposed the Omni-hardness-aware framework, which can adapt across various visual tactile sensors and analyzed factors affecting the generalization of hardness perception; regulated network learning through the Light Balance Module and Force Scale Module, verified the cross-sensor transferability of the learned representations, and demonstrated potential applications in downstream tasks such as natural object perception, tumor detection, and grasp stability prediction.

ExploRLLM: Guiding Exploration in Reinforcement Learning with Large Language Models

Runyu Ma, J. Kober

Robotic IntelligenceTransformerLarge Language ModelReinforcement Learning

🎯 What it does: Proposed an exploration method called ExploRLLM that combines large language models and reinforcement learning for robotic grasping and manipulation.

Express Yourself: Enabling Large-Scale Public Events Involving Multi-Human-Swarm Interaction for Social Applications with MOSAIX

Merihan Alhafnawi, Sabine Hauert

Robotic Intelligence

🎯 What it does: Conducting multi-person and group interactive activities using MOSAIX swarm robots in museums, collecting and aggregating public input, and providing visualization tools.

Extended Friction Models for the Physics Simulation of Servo Actuators

Marc Duclusaud, Olivier Ly

Time SeriesPhysics Related

🎯 What it does: An extended friction model was proposed to more accurately simulate the dynamics of servo actuators. Verification was achieved through comprehensive analysis of multiple friction models, parameter identification using trajectory data recorded by a pendulum test rig, and integration of the model into a physics engine, validated on four servo actuators and a 2R manipulator.

FACET: Fast and Accurate Event-Based Eye Tracking Using Ellipse Modeling for Extended Reality

Junyuan Ding, Qinyu Chen

Pose EstimationOptimizationComputational EfficiencyConvolutional Neural NetworkTime Series

🎯 What it does: Developed a gaze tracking system called FACET based on event cameras, which directly outputs pupil ellipse parameters from event data.

Fan-Out Revisited: The Impact of the Human Element on Scalability of Human Multi-Robot Teams

L. Perkins, Michael A. Goodrich

Robotic Intelligence

🎯 What it does: Propose a new fan-out model and verify it through experimental studies

Fast and Accurate Task Planning using Neuro-Symbolic Language Models and Multi-Level Goal Decomposition

Minseo Kwon, Young J. Kim

Robotic IntelligenceTransformerLarge Language Model

🎯 What it does: Proposed a neuro-symbolic task planner that uses LLM to decompose complex tasks into subgoals and employs symbolic planning or LLM-based planning with MCTS depending on subgoal complexity;

Fast Global Localization on Neural Radiance Field

Mangyu Kong, Euntai Kim

Pose EstimationNeural Radiance FieldSimultaneous Localization and Mapping

🎯 What it does: Performing global localization on NeRF maps, proposing the Fast Loc-NeRF method to accelerate localization

Fast LiDAR Data Generation with Rectified Flows

Kazuto Nakashima, Ryo Kurazume

GenerationData SynthesisAutonomous DrivingTransformerRectified FlowPoint Cloud

🎯 What it does: Propose a fast high-fidelity LiDAR generation model R2Flow based on linear trajectory learning, and design an efficient Transformer architecture to handle LiDAR's distance and reflectance image representations.

Fast Online Learning of CLiFF-Maps in Changing Environments

Yufei Zhu, Martin Magnusson

Computational EfficiencySimultaneous Localization and Mapping

🎯 What it does: Proposes an online update method for CLiFF-Map that can actively detect and adapt to changes in human flow patterns while preserving historical movement patterns;

Fast Policy Synthesis with Variable Noise Diffusion Models

Sigmund H. Høeg, Olav Egeland

Reinforcement LearningDiffusion model

🎯 What it does: Proposed Streaming Diffusion Policy (SDP), which achieves fast policy synthesis by generating partially denoised action trajectories with variable noise levels;

Fault Management System for the Safety of Perception Systems in Highly Automated Agricultural Machines

Changjoo Lee, Timo Oksanen

Safty and PrivacyAgriculture Related

🎯 What it does: Proposed a fault management system (FMS) and developed an improved image quality safety model (IQSM) for detecting, diagnosing, and mitigating risks in environmental perception systems.

FDPP: Fine-Tune Diffusion Policy with Human Preference

Yuxin Chen, Diego Romeres

Reinforcement Learning from Human FeedbackReinforcement LearningDiffusion model

🎯 What it does: Proposes the FDPP method, which learns a reward function based on human preferences and then fine-tunes the pre-trained diffusion policy through reinforcement learning to balance new preferences with the original task.

Feasibility-Aware Imitation Learning from Observations Through a Hand-Mounted Demonstration Interface

Keiichiro Takahashi, Takamitsu Matsubara

Representation LearningRobotic IntelligenceImage

🎯 What it does: Proposed a feasibility-oriented observation-based behavior cloning method called FABCO, which learns robot strategies from human demonstrations and uses visual feedback to help demonstrators improve their demonstrations.

Feature Extractor or Decision Maker: Rethinking the Role of Visual Encoders in Visuomotor Policies

Ruiyu Wang, Florian T. Pokorny

Representation LearningRobotic Intelligence

🎯 What it does: Proposed a visual alignment test framework to evaluate the effectiveness of functional separation in visual encoders under end-to-end and pre-training strategies, and study their contribution to decision-making.

FedDet: Data Poisoning Attack Detection for Federated Skeleton-based Action Recognition

Min Hyuk Kim, S. Yoo

RecognitionFederated LearningAdversarial AttackGraph

🎯 What it does: Propose FedDet, a method for detecting data poisoning attacks in federated skeleton action recognition, and develop a prototype-based attack detector.

FedEFM: Federated Endovascular Foundation Model with Unseen Data

Tuong Khanh Long Do, Anh Nguyen

Federated LearningKnowledge DistillationRepresentation LearningBiomedical Data

🎯 What it does: Propose training foundational models for intravascular catheters and guidewires in a decentralized federated learning environment, using differentiable Earth Mover's Distance to address unseen data issues.

Feedback RoI Features Improve Aerial Object Detection

Botao Ren, Zhidong Deng

Object DetectionConvolutional Neural NetworkImage

🎯 What it does: Propose a Feedback Multi-layer Feature Extractor (Flex) that dynamically adjusts feature selection using image-level and instance-level feedback information to enhance object detection performance.

FeelAnyForce: Estimating Contact Force Feedback from Tactile Sensation for Vision-Based Tactile Sensors

A. Shahidzadeh, Y. Aloimonos

Robotic IntelligenceTransformerImageTabular

🎯 What it does: Collect 200K indentation dataset, train a multi-head Transformer model to achieve 3D contact force estimation for visual tactile sensors, and generalize across different objects and sensors.

FGO-SLAM: Enhancing Gaussian SLAM with Globally Consistent Opacity Radiance Field

Fan Zhu, Hui Zhu

Autonomous DrivingOptimizationNeural Radiance FieldGaussian SplattingSimultaneous Localization and MappingPoint CloudMesh

🎯 What it does: Proposes FGO-SLAM, a Gaussian SLAM system that enhances high-quality geometric mapping by leveraging a globally consistent opaque radiance field, achieving more robust tracking and mapping through global pose adjustment, depth distortion and normal consistency terms, as well as isosurface extraction based on triangular meshes.

Fine-Grained Open-Vocabulary Object Detection with Fined-Grained Prompts: Task, Dataset and Benchmark

Ying Liu, Tengqi Ye

Object DetectionPrompt EngineeringImageBenchmark

🎯 What it does: This paper proposes the 3F-OVD task, extending supervised fine-grained object detection to open-vocabulary scenarios, creating the NEU-171K fine-grained dataset, benchmarking existing state-of-the-art object detectors on this dataset, and proposing a simple and effective post-processing technique.

Fine-Tuning Hard-to-Simulate Objectives for Quadruped Locomotion: A Case Study on Total Power Saving

Ruiqian Nai, Yang Gao

OptimizationRobotic IntelligenceReinforcement LearningWorld Model

🎯 What it does: Propose a data-driven framework based on real-world data to fine-tune quadruped robot gait policies, optimizing hard-to-simulate objectives (e.g., energy consumption)

Finite-Step Capturability and Recursive Feasibility for Bipedal Walking in Constrained Regions

Shubham S. Kumbhar, Ioannis Poulakakis

OptimizationRobotic Intelligence

🎯 What it does: Propose a model predictive control (MPC) formulation based on the linear inverted pendulum (LIP) model for bipedal gait planning, ensuring recursive feasibility within constrained regions.

FisheyeDepth: A Real Scale Self-Supervised Depth Estimation Model for Fisheye Camera

Guoyang Zhao, Jun Ma

Depth EstimationImage

🎯 What it does: Proposed a self-supervised depth estimation model called FisheyeDepth for fisheye cameras, integrating the fisheye camera model with real-scale pose and adopting a multi-channel output strategy to enhance robustness.

FitnessAgent: A Unified Agent Framework for Open-Set and Personalized Fitness Evaluation

Zhenhui Tang, Peng Wang

TransformerLarge Language ModelAgentic AIBiomedical Data

🎯 What it does: Proposes the FitnessAgent framework, capable of performing unpretrained evaluation for any movement in open-set and personalized fitness assessment scenarios;

FLAF: Focal Line and Feature-Constrained Active View Planning for Visual Teach and Repeat

Changfei Fu, Hong Zhang

OptimizationRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposes an active view planning method called FLAF based on focus lines and feature constraints, used to automatically adjust the orientation of a rotating camera in mobile robot navigation.

Flapping-Wing Flying Robot with Integrated Dual-Arm Scissors-Type Flora Sampling System

Rodrigo Gordillo Durán, A. Ollero

Safty and PrivacyRobotic Intelligence

🎯 What it does: Integrate an improved dual-arm scissor sampling system onto a flapping-wing robot, achieving safe cutting functionality

FLaRe: Achieving Masterful and Adaptive Robot Policies with Large-Scale Reinforcement Learning Fine-Tuning

Jiaheng Hu, Kiana Ehsan

Robotic IntelligenceSupervised Fine-TuningReinforcement Learning

🎯 What it does: Proposes the FLaRe framework, which leverages large-scale reinforcement learning fine-tuning methods to enhance the performance of pre-trained robot policies

Flat'n'Fold: A Diverse Multi-Modal Dataset for Garment Perception and Manipulation

Lipeng Zhuang, Gerardo Aragon-Camarasa

Robotic IntelligenceMultimodalityPoint CloudBenchmark

🎯 What it does: Created and released a large-scale garment manipulation dataset called Flat'n'Fold, containing 1,212 human and 887 robot examples of complete garment folding processes, with multi-view RGB-D images, point clouds, and action data.

FlatFusion: Delving Into Details of Sparse Transformer-Based Camera-LiDAR Fusion for Autonomous Driving

Yutao Zhu, Junchi Yan

Autonomous DrivingTransformerImageMultimodalityPoint Cloud

🎯 What it does: Addressing sparse Transformer visual-LiDAR fusion, systematically explore and integrate various design schemes, proposing the FlatFusion framework.

FLEX: A Framework for Learning Robot-Agnostic Force-Based Skills Involving Sustained Contact Object Manipulation

Shijie Fang, Jivko Sinapov

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a robot-agnostic, force-based continuous contact object manipulation framework called FLEX, which simplifies the action space by directly applying forces to the object's surface;

FlightForge: Advancing UAV Research with Procedural Generation of High-Fidelity Simulation and Integrated Autonomy

David Capek, M. Saska

Autonomous DrivingRobotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Propose FlightForge UAV open-source simulator, providing advanced rendering, rich control methods, and implementing procedural environment generation with a complete autonomous navigation system integration, supporting long-distance flight in complex unknown environments.

Flora: Sample-Efficient Preference-Based Rl Via Low-Rank Style Adaptation of Reward Functions

Daniel Marta, Iolanda Leite

Robotic IntelligenceReinforcement Learning from Human FeedbackSupervised Fine-TuningReinforcement Learning

🎯 What it does: Propose a low-rank parameter reward function adaptation method to regulate pre-trained robot behavior through sample-efficient preference-based reinforcement learning, balancing the original task and user preferences;

Flying Quadrotors in Tight Formations Using Learning-Based Model Predictive Control

K. Y. Chee, M. Hsieh

OptimizationRobotic IntelligenceTime Series

🎯 What it does: Propose a framework that combines physics-based modeling with a lightweight data-driven model to capture complex aerodynamic effects in tightly formation quadrotor flights, and integrate this model into a learning-based nonlinear model predictive controller (MPC).

Flying Through Moving Gates without Full State Estimation

Ralf Römer, Angela P. Schoellig

Autonomous DrivingOptimizationRobotic IntelligenceImage

🎯 What it does: Proposes a control algorithm that does not rely on track maps or VIO, utilizing only monocular visual measurements to navigate through doors accurately in environments where doors are moving or there is wind.

FlyKites: Human-Centric Interactive Exploration and Assistance Under Limited Communication

Yuyang Zhang, Meng Guo

OptimizationRobotic Intelligence

🎯 What it does: Proposes a multi-robot system called FlyKites for human-robot interactive exploration and assistance in environments with human-constrained communication.

Focused Blind Switching Manipulation Based on Constrained and Regional Touch States of Multi-Fingered Hand Using Deep Learning

S. Funabashi, Tetsuya Ogata

Robotic IntelligenceRecurrent Neural NetworkAuto EncoderMultimodality

🎯 What it does: This study proposes a multi-fingered hand depth learning blind switching manipulation method based on constrained and regional tactile states, designing a loss function constrained by tactile states and an attention mechanism, and achieving motion switching under different tactile states through the AE-LSTM model, with the verification task being bottle cap opening.

FogROS2-PLR: Probabilistic Latency-Reliability for Cloud Robotics

Kai-Peng Chen, Kenneth Y. Goldberg

OptimizationRobotic Intelligence

🎯 What it does: This paper proposes the FogROS2-PLR framework, which utilizes multi-path network interfaces and cloud server redundancy to enhance the latency reliability of cloud robotics systems.

Force Admittance Control of an Underactuated Gripper with Full-State Feedback

Chunpeng Wang, M. A. Estrada

Robotic Intelligence

🎯 What it does: A scheme for impedance control and fingertip contact detection of a linkage gripper driven by a pneumatic rolling diaphragm actuator was developed, with modeling of the linkage and fluid transmission completed. The verification demonstrated the ability to regulate grip force through impedance control (RMS error below 0.5N) and to locate contact points on the linkage. The gripper was also shown to perform obstacle-free blind search and detect object loss.

Force Myography Based Torque Estimation in Human Knee and Ankle Joints

Charlotte Marquardt, Tamim Asfour

Biomedical Data

🎯 What it does: Propose a knee and ankle torque estimation method based on FMG, combining muscle activity data with joint angles and velocities, and using Gaussian process regression to learn the model.

Force-Conditioned Diffusion Policies for Compliant Sheet Separation Tasks in Bimanual Robotic Cells

Rishabh Shukla, Satyandra K. Gupta

Robotic IntelligenceDiffusion model

🎯 What it does: Collect human demonstrations via motion capture, train a diffusion-based force perception strategy, enabling dual-arm collaborative robots to apply precise forces and separate adhesive films.

ForceMimic: Force-Centric Imitation Learning with Force-Motion Capture System for Contact-Rich Manipulation

Wenhai Liu, Cewu Lu

Robotic IntelligenceAgriculture Related

🎯 What it does: Proposed the ForceMimic system, which includes a robot-agnostic ForceCapture demonstration collection system and the HybridIL hybrid force-kinematic imitation learning algorithm, with experimental validation on a vegetable peeling task.

Foresee and Act Ahead: Task Prediction and Pre-Scheduling Enabled Efficient Robotic Warehousing

Bo Cao, Hesheng Wang

OptimizationRobotic IntelligenceGraph Neural Network

🎯 What it does: Proposed a pre-scheduling enhanced warehouse framework that includes task flow prediction and hybrid task allocation.

Forward Invariance in Trajectory Spaces for Safety-Critical Control

M. Vahs, Jana Tumova

Optimization

🎯 What it does: Proposes a control method that ensures safety by achieving forward invariance in the trajectory space

Fostering Trust Through Gesture and Voice-Controlled Robot Trajectories in Industrial Human-Robot Collaboration*

Giulio Campagna, Matthias Rehm

RecognitionRobotic IntelligenceVision-Language-Action ModelMultimodality

🎯 What it does: Designed and evaluated a framework that utilizes human gestures and voice commands to control robot trajectories, aiming to enhance trust and efficiency in industrial human-robot collaboration.

Foundation Feature-Driven Online End-Effector Pose Estimation: A Marker-Free and Learning-Free Approach

Tianshu Wu, Hao Dong

Pose EstimationRobotic IntelligenceImage

🎯 What it does: Propose an online, markerless, and learning-free end-effector pose estimation algorithm called FEEPE, which utilizes pre-trained visual features for 2D-3D correspondence and achieves 6D pose estimation through PnP, while resolving ambiguities caused by partial observations and symmetry via multi-historical keyframe optimization;

Foundation Models for Rapid Autonomy Validation

Alec Farid, Christoffer Heckman

Autonomous DrivingRepresentation LearningSupervised Fine-TuningAuto Encoder

🎯 What it does: Reconstruct driving scenarios using a behavior-based model (MAE), and rapidly assess collision rates and severity through embedding space clustering and fine-tuning difficulty labels.

FRASA: An End-to-End Reinforcement Learning Agent for Fall Recovery and Stand Up of Humanoid Robots

Clément Gaspard, Olivier Ly

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed FRASA, a deep reinforcement learning agent that integrates fall recovery and stand-up strategies into a unified framework;

FreeDriveRF: Monocular RGB Dynamic NeRF Without Poses for Autonomous Driving via Point-Level Dynamic-Static Decoupling

Yue Wen, Hesheng Wang

Autonomous DrivingOptimizationNeural Radiance FieldOptical FlowImage

🎯 What it does: Construct a dynamic driving scene NeRF using only continuous RGB images without pose input, employing point-level dynamic-static decoupling and optical flow constraints.

From Ceilings to Walls: Universal Dynamic Perching of Quadrotors on Surfaces with Variable Orientations

Bryan Habas, Bo Cheng

Robotic IntelligenceReinforcement Learning

🎯 What it does: A non-dimensional framework and deep reinforcement learning were used to systematically evaluate the dynamic hover capability of quadrotors of different sizes on surfaces with various inclination angles. Experimental validation demonstrated consistent hover performance when maintaining geometric proportions. Additionally, the study investigated the impact of joint stiffness and damping of the landing gear on hover behavior, revealing a critical velocity threshold that determines successful hover.

From Cognition to Precognition: A Future-Aware Framework for Social Navigation

Zeying Gong, Junwei Liang

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: Propose a reinforcement learning architecture named Falcon for social navigation, which explicitly predicts human trajectories and penalizes actions that obstruct future paths, while creating a new SocialNav evaluation benchmark and two datasets.

From Configuration-Space Clearance to Feature-Space Margin: Sample Complexity in Learning-Based Collision Detection

Sapir Tubul, Oren Salzman

Robotic Intelligence

🎯 What it does: By analyzing the sample complexity of SVM classifiers, a collision detection algorithm that provides statistical error guarantees is proposed.

From Imitation to Refinement - Residual Rl for Precise Assembly

Lars Ankile, Pulkit Agrawal

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose a new method called Resip, combining behavior cloning with residual strategies in reinforcement learning to enhance the reliability of robots in high-precision assembly tasks.

From Simple to Complex Skills: The Case of In-Hand Object Reorientation

Haozhi Qi, Jitendra Malik

Pose EstimationRobotic IntelligenceReinforcement Learning

🎯 What it does: Construct a hierarchical strategy using low-level rotation skills to achieve in-hand object reorientation, and employ a general pose estimator based on proprioception, low-level skill prediction, and control error.

Fuel-Optimal Operational Speed Planning for Autonomous Trucking on Highways

Wei Li, Ruigang Yang

Autonomous DrivingOptimizationTime Series

🎯 What it does: Defined the end-to-end speed planning problem (OSP) and developed a dataset containing 220,000 km of real-world data from 400 trucks, a closed-loop simulator, and a baseline speed planning method based on dynamic programming.

Full-Order Sampling-Based MPC for Torque-Level Locomotion Control via Diffusion-Style Annealing

Haoru Xue, Guanya Shi

OptimizationDiffusion model

🎯 What it does: Proposed a sampling-based MPC framework called DIAL-MPC, which achieves real-time joint-level control under a full-order dynamic model using a diffusion annealing process

Fully Differentiable Adaptive Informative Path Planning

Kalvik Jakkala, Srinivas Akella

OptimizationRobotic Intelligence

🎯 What it does: Proposed a fully differentiable adaptive information path planning algorithm for rapidly and efficiently gathering environmental information.

Fusionsense: Bridging Common Sense, Vision, and Touch for Robust Sparse-View Reconstruction

Irving Fang, Jing Zhang

Robotic IntelligenceGaussian SplattingMultimodality

🎯 What it does: Introduce the FusionSense framework to enable robots to perform 3D reconstruction by fusing foundational model priors under sparse visual and tactile observations.

FutureNet-LoF: Joint Trajectory Prediction and Lane Occupancy Field Prediction with Future Context Encoding

Mingkun Wang, Wenjing Yang

Autonomous DrivingSequentialBenchmark

🎯 What it does: Propose the FutureNet and LOF frameworks, which utilize trajectory prediction to encode future scenarios and achieve joint prediction of trajectories and lane occupancy fields.

FuzzRisk: Online Collision Risk Estimation for Autonomous Vehicles Based on Depth-Aware Object Detection via Fuzzy Inference

Brian Hsuan-Cheng Liao, Alois Knoll

Object DetectionDepth EstimationAutonomous DrivingPoint Cloud

🎯 What it does: Propose an online collision risk estimation framework that infers collision risks for autonomous vehicles by leveraging inconsistencies between depth-aware 2.5D object detection and traditional 3D object detection.

GA-TEB: Goal-Adaptive Framework for Efficient Navigation Based on Goal Lines

Qianyi Zhang, Jingtai Liu

Autonomous Driving

🎯 What it does: Proposed the GA-TEB framework based on target lines for efficient obstacle avoidance navigation.

GAGrasp: Geometric Algebra Diffusion for Dexterous Grasping

Tao Zhong, Christine Allen-Blanchette

Robotic IntelligenceDiffusion model

🎯 What it does: Proposed the GAGrasp framework, which utilizes geometric algebra representation to achieve SE(3) equivariance, generates feasible and stable grasping postures, and incorporates a differentiable physics-informed refinement layer;

GAPartManip: A Large-Scale Part-Centric Dataset for Material-Agnostic Articulated Object Manipulation

Wenbo Cui, He Wang

Pose EstimationDepth EstimationRobotic IntelligenceImageBenchmark

🎯 What it does: Proposed and constructed a large-scale, part-centric articulated object manipulation dataset called GAPartManip, conducted integrated experiments with various state-of-the-art depth estimation and interactive pose prediction methods, and further introduced a modular framework to achieve more robust and generalizable manipulation performance.

GARAD-SLAM: 3D Gaussian Splatting for Real-Time Anti Dynamic SLAM

Mingrui Li, Hongyu Wang

Gaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Proposed GARAD-SLAM, a real-time 3D Gaussian Splatting (3DGS) SLAM system for dynamic scenes, capable of dynamically segmenting Gaussian points and avoiding false deletions through rendering penalties;

GaRLIO: Gravity Enhanced Radar-LiDAR-Inertial Odometry

Chiyun Noh, Ayoung Kim

Pose EstimationAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a gravity-enhanced trajectory estimation method called GaRLIO, which integrates radar, LiDAR, and inertial measurement unit (IMU) data. The method leverages direct velocity information from radar to improve gravity estimation accuracy and reduces vertical drift by eliminating dynamic objects using radar data.

Gassidy: Gaussian Splatting SLAM in Dynamic Environments

Long Wen, A. Knoll

Gaussian SplattingSimultaneous Localization and MappingImage

🎯 What it does: Developed an RGB-D dense SLAM method called Gaussian Splatting SLAM in Dynamic Environments (Gassidy).

Gate-Aware Online Planning for Two-Player Autonomous Drone Racing

Fangguo Zhao, Shuo Li

Autonomous DrivingOptimization

🎯 What it does: Developed a system called Pairwise Model Predictive Control (PMPC), which can guide two drones through gates at opposite directions in a race track in the shortest time without collision, and validated its feasibility through simulation and real experiments.

Gaussian Splatting Visual MPC for Granular Media Manipulation

Wei-Cheng Tseng, Florian Shkurti

OptimizationRobotic IntelligenceGaussian SplattingImage

🎯 What it does: Propose a visual dynamics model based on Gaussian splatting representation, combined with model predictive control (MPC) for manipulating granular media.

Gaussian-LIC: Real-Time Photo-Realistic SLAM with Gaussian Splatting and LiDAR-Inertial-Camera Fusion

Xiaolei Lang, Xingxing Zuo

Autonomous DrivingOptimizationComputational EfficiencyGaussian SplattingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Propose a real-time realistic SLAM method based on Gaussian Splatting and LiDAR-IMU-Camera fusion.

GaussianRoom: Improving 3D Gaussian Splatting with SDF Guidance and Monocular Cues for Indoor Scene Reconstruction

Haodong Xiang, Xue-ping Liu

GenerationOptimizationGaussian SplattingPoint Cloud

🎯 What it does: Propose a unified optimization framework that combines neural SDF with 3D Gaussian expansion to achieve accurate geometry reconstruction and real-time rendering.

Gaze and Go: Harnessing Visual Attention Valence in Upper-Limb Robotic Rehabilitation With Tailored Gamification and Eye Tracking for Neuroplasticity

Daomiao Wang, Hongliu Yu

Robotic IntelligenceBiomedical Data

🎯 What it does: Developed the ArmGuider Pro upper limb rehabilitation robot system, integrating hand-eye collaboration and gaze-triggered assistive features into serious games designed for rehabilitation.

GazeHTA: End-to-End Gaze Target Detection with Head-Target Association

Zhi-Yi Lin, Xucong Zhang

Object DetectionDiffusion modelImage

🎯 What it does: Propose an end-to-end multi-person gaze target detection framework called GazeHTA, which directly predicts head-target instance connections from scene images.

Gen-Drive: Enhancing Diffusion Generative Driving Policies with Reward Modeling and Reinforcement Learning Fine-Tuning

Zhiyu Huang, Chen Lv

Autonomous DrivingReinforcement LearningVision Language ModelDiffusion model

🎯 What it does: Propose the Gen-Drive framework, which transforms traditional prediction with deterministic planning into a generative-evaluation planning approach. It uses a behavior diffusion model to generate diverse future scenarios and trains a scene evaluation (reward) model and an RL fine-tuning framework to improve generation quality and planning effectiveness.

GenCo: A Dual VLM Generate-Correct Framework for Adaptive Peg-in-Hole Robotics

Zhengxue Zhou, A. I. Cooper

Robotic IntelligenceSupervised Fine-TuningMixture of ExpertsVision Language ModelMultimodality

🎯 What it does: Implemented a Generate-Correct framework based on dual Vision-Language Models (VLMs) for adaptive blind peg-in-hole robotic tasks, integrating a motion generator and a motion expert to generate and correct actions during execution.

General-Purpose Clothes Manipulation with Semantic Keypoints

Yuhong Deng, David Hsu

Robotic IntelligenceLarge Language ModelVision-Language-Action ModelImage

🎯 What it does: Propose the CLASP method, which utilizes semantic keypoints to enable robots to perform various operational tasks on different types of clothing.

Generalizable Imitation Learning Through Pre-Trained Representations

Wei-Di Chang, Gregory Dudek

Representation LearningTransformerContrastive LearningImage

🎯 What it does: Leveraging pre-trained self-supervised visual Transformer models and their semantic capabilities, the DVK algorithm is proposed to enhance the generalization ability of imitation learning in demonstration learning.