arXivSub Start free trial

ICRA 2025 Papers — Page 13

IEEE International Conference on Robotics and Automation · 1604 papers

Reliable and Efficient Multi-Agent Coordination via Graph Neural Network Variational Autoencoders

Yue Meng, Federico Pecora

OptimizationComputational EfficiencyRobotic IntelligenceGraph Neural NetworkAuto EncoderGraph

🎯 What it does: Utilizing GNN-VAE to rapidly solve large-scale multi-robot coordination problems

ReloPush: Multi-Object Rearrangement in Confined Spaces with a Nonholonomic Mobile Robot Pusher

Jeeho Ahn, Christoforos Mavrogiannis

OptimizationRobotic IntelligenceGraph

🎯 What it does: Proposes a graph-based pushing planning framework called ReloPush for non-Hungarian constrained mobile robots to rearrange multiple objects in confined spaces.

ReMEmbR: Building and Reasoning Over Long-Horizon Spatio-Temporal Memory for Robot Navigation

Abrar Anwar, Yan Chang

Robotic IntelligenceLarge Language ModelVision Language ModelVideoTextRetrieval-Augmented Generation

🎯 What it does: Built and utilized the retrieval-enhanced memory system ReMEmbR to enable question answering in long-duration robot navigation videos.

Remote: Real-Time Ego-Motion Tracking for Various Endoscopes via Multimodal Visual Feature Learning

Liangjing Shao, Xinrong Chen

Pose EstimationOptical FlowMultimodalityBiomedical Data

🎯 What it does: Proposed a new framework for real-time self-motion tracking in endoscopy

Renderworld: World Model with Self-Supervised 3D Label

Ziyang Yan, Yuexin Ma

Autonomous DrivingAuto EncoderGaussian SplattingWorld ModelImage

🎯 What it does: Propose the RenderWorld framework, a pure vision-based end-to-end autonomous driving system. It first generates 3D occupancy labels using a self-supervised Gaussian Img2Occ module, then encodes the labels with AM-VAE, and finally uses a world model for prediction and planning.

Reservoir Computing Encodes Physical Adaptations for Reinforcement Learning

Cross Giannetto, Arsen Abdulali

Robotic IntelligenceRecurrent Neural NetworkReinforcement Learning

🎯 What it does: Propose a framework that combines Reservoir Computing (RC) with the FORCE learning rule to enhance the adaptability of reinforcement learning strategies under different robot body configurations.

Resettable Land Anchor Launcher for Unmanned Rover Rescue and Slope Climbing

Aaryan Kainth, Nicholas D. Naclerio

Robotic Intelligence

🎯 What it does: Designed a resettable ground-anchored launcher integrated into a half-meter-long unmanned vehicle platform, and conducted field tests at the NASA Glenn Research Center SLOPE laboratory.

Residual Descent Differential Dynamic Game (RD3G) - A Fast Newton Solver for Constrained General Sum Games

Zhiyuan Zhang, Panagiotis Tsiotras

Optimization

🎯 What it does: Proposed a solver called RD3G based on the Newton method for finding local Nash equilibria in constrained multi-agent control problems;

Residual Policy Learning for Perceptive Quadruped Control Using Differentiable Simulation

Jingru Luo, Marco Hutter

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Propose guiding policy search through learning residual strategies in the first-order policy gradient training to enhance the final reward of quadruped robots in dense contact tasks, and demonstrate end-to-end training for global perception navigation and motion control within minutes.

Resolution Optimal Motion Planning for Medical Needle Steering from Airway Walls in the Lung

J. Hoelscher, Ron Alterovitz

OptimizationRobotic IntelligenceBiomedical Data

🎯 What it does: Developed a resolution-optimal motion planning algorithm for inserting medical needles from the airway wall into the target position in the lungs.

Retinex-BEVFormer: Using Retinex to Enhance Multi-View Image-Based BEV Detector in Low Light Scenes

Xuan Liu, Xinkai Wu

RestorationObject DetectionAutonomous DrivingTransformerImage

🎯 What it does: Propose Retinex-BEVFormer, which generates illumination information using Retinex theory for illumination-guided feature fusion, and propose the MVB-Retinex module to balance illumination estimation across multi-view images.

Retrieval-Augmented Hierarchical in-Context Reinforcement Learning and Hindsight Modular Reflections for Task Planning with LLMs

Chuanneng Sun, D. Pompili

Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningTextBenchmarkRetrieval-Augmented Generation

🎯 What it does: Proposed and implemented the Retrieval-Augmented Hierarchical in-context Reinforcement Learning (RAHL) framework, leveraging LLM for hierarchical task decomposition and subtask execution, while integrating Hindsight Modular Reflection (HMR) to enhance multi-round execution performance.

ReVLA: Reverting Visual Domain Limitation of Robotic Foundation Models

Sombit Dey, D. Paudel

Domain AdaptationRobotic IntelligenceTransformerVision-Language-Action Model

🎯 What it does: Investigate the visual generalization capabilities of existing robotic foundation models, propose an evaluation framework, and recover visual OOD generalization through progressive backbone inversion (model merging) technology, resulting in the ReVLA model.

RGB-Only Gaussian Splatting SLAM for Unbounded Outdoor Scenes

Sicheng Yu, Hao Wang

OptimizationGaussian SplattingSimultaneous Localization and MappingImage

🎯 What it does: Proposed and implemented an RGB-only Gaussian splatting SLAM method for unbounded outdoor scenes called OpenGS-SLAM;

RINA: Rapid Introspective Neural Adaptation for Out-of-Distribution Payload Configurations on Quadruped Robots

Oscar Youngquist, Hao Zhang

Robotic Intelligence

🎯 What it does: Proposed and implemented a Fast Self-aware Neural Adaptation (RINA) method for quadruped robots to rapidly compensate for joint torque discrepancies when carrying Out-of-Distribution (O.O.D.) payloads, enabling real-time adaptation.

RipGAN: A GAN-Based Rip Current Data Augmentation Method

Shenyang Qian, Yang Song

Object DetectionData SynthesisConvolutional Neural NetworkGenerative Adversarial NetworkImage

🎯 What it does: Proposes the RipGAN data augmentation method based on GAN, which includes a texture generator and an FFF-Unet-based Rip generator to generate realistic wave textures and surge images, thereby enriching training data.

RISED: Accurate and Efficient RGB-Colorized Mapping Using Image Selection and Point Cloud Densification

Changjian Jiang, Yu Zhang

Depth EstimationAutonomous DrivingComputational EfficiencySimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposed the RISED system, which achieves RGB-color mapping through image selection and point cloud densification.

Risk-Averse Model Predictive Control for Racing in Adverse Conditions

T. Lew, John K. Subosits

Autonomous DrivingOptimization

🎯 What it does: Designed and implemented a risk-averse MPC framework for racing control under adverse conditions.

Risk-Aware Energy-Constrained UAV-UGV Cooperative Routing Using Attention-Guided Reinforcement Learning

Md Safwan Mondal, Pranav A. Bhounsule

Autonomous DrivingOptimizationTransformerReinforcement Learning

🎯 What it does: Proposes a risk-aware deep reinforcement learning framework that employs an encoder-decoder Transformer model with a multi-head attention mechanism to address UAV-UGV cooperative path planning problems, minimizing task time while meeting energy risk thresholds.

Risk-Aware Integrated Task and Motion Planning for Versatile Snake Robots Under Localization Failures

A. Jasour, Rohan Thakker

OptimizationRobotic Intelligence

🎯 What it does: Proposes the BLISS method, combining blind walking with intermittent scanning to address localization and collision risks for serpentine robots in extreme terrains and confined environments.

RL-GSBridge: 3D Gaussian Splatting Based Real2Sim2Real Method for Robotic Manipulation Learning

Yuxuan Wu, Hesheng Wang

Robotic IntelligenceReinforcement LearningGaussian SplattingMesh

🎯 What it does: Proposed a Real2Sim2Real framework RL-GSBridge based on 3D Gaussian expansion, integrating 3D GS into the RL simulation pipeline to achieve zero-shot visual depth reinforcement learning Sim2Real transfer.

RL-OGM-Parking: Lidar OGM-Based Hybrid Reinforcement Learning Planner for Autonomous Parking

Zhitao Wang, Ming Yang

Autonomous DrivingReinforcement LearningSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposed a hybrid planner that combines rule-based planning based on Reeds-Shepp and learning-based planning based on reinforcement learning, using real-time LiDAR occupancy grid map (OGM) representation to achieve seamless transfer from simulation to reality.

RLCNet: A Novel Deep Feature-Matching-Based Method for Online Target-Free Radar-LiDAR Calibration

Kai-Rui Luan, Huimin Lu

Pose EstimationAutonomous DrivingPoint Cloud

🎯 What it does: Propose an online target-free radar-LiDAR extrinsic calibration method based on deep feature matching, formulating the calibration problem as a cross-modal point cloud registration task, first performing keypoint matching and then refining with dense matching.

RLPP: A Residual Method for Zero-Shot Real-World Autonomous Racing on Scaled Platforms

Edoardo Ghignone, Michele Magno

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose an RLPP framework that combines the Pure Pursuit (PP) controller with a residual module based on reinforcement learning (RL) for real-world large-scale autonomous racing without requiring zero-shot learning.

RM-Planner: Integrating Reinforcement Learning with Whole-Body Model Predictive Control for Mobile Manipulation

Zixuan Zhuang, Hui Cheng

Robotic IntelligenceReinforcement LearningPoint Cloud

🎯 What it does: Propose a planning method for mobile manipulation in unknown complex environments, integrating reinforcement learning with full-body MPC.

RM4D: A Combined Reachability and Inverse Reachability Map for Common 6-/7-Axis Robot Arms by Dimensionality Reduction to 4D

Martin Rudorfer

Robotic Intelligence

🎯 What it does: Propose a Reachability Map 4D (RM4D) that reduces the workspace of common 6/7-axis robotic arms from 6D to 4D, enabling both forward and inverse queries with a single 4D data structure, and demonstrate its application in scenarios for finding suitable robot base positions.

RMP-YOLO: A Robust Motion Predictor for Partially Observable Scenarios Even if You Only Look Once

Jiawei Sun, Marcelo H. Ang

Autonomous DrivingTransformerBenchmark

🎯 What it does: Propose the RMP-YOLO unified framework, which achieves robust motion prediction in partially observable scenarios by first reconstructing complete historical trajectories.

RMSeg-UDA: Unsupervised Domain Adaptation for Road Marking Segmentation Under Adverse Conditions

Yi-Chang Cai, Chiao-Tung Chan

SegmentationDomain AdaptationAutonomous DrivingGenerative Adversarial NetworkImage

🎯 What it does: Proposed an unsupervised domain adaptation framework for road sign segmentation called RMSeg-UDA, combining schedule self-training and class-conditioned adversarial training, leveraging labeled normal weather data and unlabeled data from other domains to train the model.

Robi Butler: Multimodal Remote Interaction with a Household Robot Assistant

Anxing Xiao, David Hsu

Robotic IntelligenceTransformerLarge Language ModelVision Language ModelMultimodality

🎯 What it does: Developed the Robi Butler robot assistant to enable multimodal remote interaction, supporting visual, speech, and gesture commands, using LLM to generate multi-step action plans.

RoBiFusion: A Robust and Bidirectional Interaction Camera-LiDAR 3D Object Detection Framework

Xubin Wen, Si-Yu Xia

Object DetectionAutonomous DrivingTransformerImageMultimodalityPoint Cloud

🎯 What it does: Proposed the RoBiFusion framework to address the noise problem caused by shallow feature interactions in camera-radar 3D detection;

Robo-DM: Data Management for Large Robot Datasets

Kai-Peng Chen, Kenneth Y. Goldberg

CompressionData-Centric LearningRobotic IntelligenceTransformerVideo

🎯 What it does: Proposed Robo-DM, a cloud-based, efficient open-source data management tool for collecting, sharing, and learning robot data, which stores data in EBML format and significantly reduces data volume, transmission cost, and loading time through compression and memory-mapped decoding cache.

Robo-GS: A Physics Consistent Spatial-Temporal Model for Robotic Arm with Hybrid Representation

Haozhe Lou, Yongliang Shi

Robotic IntelligenceGaussian SplattingMeshPhysics Related

🎯 What it does: Proposes a Real2Sim pipeline and a hybrid representation model that integrates mesh geometry, 3D Gaussian kernels, and physical properties to generate high-fidelity digital assets for robotic arm simulations.

Robo-MUTUAL: Robotic Multimodal Task Specification via Unimodal Learning

Jianxiong Li, Xianyuan Zhan

Robotic IntelligenceVision-Language-Action ModelTextMultimodalityBenchmark

🎯 What it does: Train robots using rich single-modal instructions, pretrain a multimodal encoder, and bridge modality gaps through Collapse and Corrupt operations, enabling robots to perform multimodal task specifications in a cross-modal aligned latent space.

RoboCrowd: Scaling Robot Data Collection Through Crowdsourcing

Suvir Mirchandani, Dorsa Sadigh

Robotic IntelligenceSupervised Fine-Tuning

🎯 What it does: Proposed and implemented RoboCrowd—a new paradigm for scaling the collection of robot demonstration data by leveraging crowdsourcing principles and incentive design—and conducted a two-week field experiment in a university café, collecting over 800 interaction data points from more than 200 volunteers;

Robot Local Planner: A Periodic Sampling-Based Motion Planner with Minimal Waypoints for Home Environments

Keisuke Takeshita, Takashi Yamamoto

Robotic Intelligence

🎯 What it does: Proposes a periodic sampling whole-body trajectory planning method (Robot Local Planner) to achieve fast and safe manipulation tasks in home environments.

Robot Manipulation in Salient Vision Through Referring Image Segmentation and Geometric Constraints

Chen Jiang, Martin Jägersand

SegmentationComputational EfficiencyRobotic IntelligenceVision-Language-Action ModelImageText

🎯 What it does: Integrate a lightweight referential image segmentation model into a robot perception module in real environments to enable robot operations under language contexts;

Robot Navigation in Unknown and Cluttered Workspace with Dynamical System Modulation in Starshaped Roadmap

Kai Chen, Jun Ma

Robotic Intelligence

🎯 What it does: Proposes a motion planning and control framework for robot navigation in unknown and crowded environments based on dynamically constructing a star-shaped road network, generating a star-shaped free space representation using real-time sensor data and building an incremental connected graph, and achieving safe and smooth real-time control through dynamic system modulation (DSM).

Robot Planning Under Uncertainty for Object Assembly and Troubleshooting Using Human Causal Models

Semanti Basu, R. I. Bahar

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Integrates human causal models obtained through crowdsourcing into a collaborative robot decision framework for object assembly and fault diagnosis tasks under partially observable conditions.

Robot Policy Transfer with Online Demonstrations: An Active Reinforcement Learning Approach

Muhan Hou, Kim Baraka

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed an active learning-based online demonstration method to optimize query timing and content under a limited demonstration budget, thereby improving the success rate and sample efficiency of robot policy transfer

Robot Utility Models: General Policies for Zero-Shot Deployment in New Environments

Haritheja Etukuru, Nur Muhammad (Mahi) Shafiullah

Robotic IntelligenceLarge Language ModelMultimodality

🎯 What it does: Trained and deployed robot utility models (RUMs) capable of zero-shot direct operation in new environments, developed tools for rapidly collecting data on mobile manipulation tasks, integrated data using multimodal imitation learning, achieved on-device deployment on the Hello Robot Stretch robot, paired with an external large language model (mLLM) validator for retries, trained five utility models for tasks including opening doors, opening drawers, taking towels, taking paper bags, and repositioning fallen objects, achieving an average success rate of 90% on unseen new environments and objects.

Robot-Based Automatic Charging for Electric Vehicles Using Incremental Learning and Biomimetic Control

Chao Zeng, Chenguang Yang

Robotic Intelligence

🎯 What it does: Propose an incremental learning method based on a generalized learning system for locating charging ports in electric vehicle robots during automatic charging, and design a bio-impedance controller to achieve adaptive compliance behavior during the plug-and-unplug process, with experimental validation on UR robots.

Robotic 3D Flower Pose Estimation for Small-Scale Urban Farms

Harsh Muriki, Ai-Ping Hu

Pose EstimationRobotic IntelligenceConvolutional Neural NetworkImagePoint CloudAgriculture Related

🎯 What it does: Using FarmBot and a custom camera to acquire 3D point clouds and estimate the pose of strawberry plant flowers; by translating the occupied grid along orthogonal axes to generate six perspectives of 2D images, utilizing 2D object detection to extract flower point clouds, and then fitting three shapes (superellipse, paraboloid, and plane) to determine the pose.

Robotic Colonoscopy: Can High Fidelity Simulation Optimize Robot Design and Validation?

M. Evans, S. Dogramadzi

OptimizationRobotic IntelligenceBiomedical Data

🎯 What it does: This paper constructs an Ansys finite element simulation model to simulate the motion of the prototype robotic colonoscope on different colon surfaces, and compares the simulation results with experimental results to verify the effectiveness of the simulation method.

Robotic Dry-Stacking of Clocháin with Irregular Stones

Yifang Liu, Nils Napp

Robotic Intelligence

🎯 What it does: Studied the dry stacking construction of self-standing clocháin (stone hut) structures by robots on irregular stones, and proposed a stacking measurement method to assist in stone selection, significantly enhancing structural stability and achieving physical demonstrations in experiments.

Robotic Flexible Magnetic Retractor for Dynamic Tissue Manipulation in Endoscopic Submucosal Dissection

Wai Shing Chan, Zheng Li

Robotic Intelligence

🎯 What it does: A robotic flexible magnetic manipulator was developed that can be inserted into the lesion site through the endoscope instrument channel without removing the endoscope, for tissue manipulation during ESD surgery.

Robotic Framework for Iterative and Adaptive Profile Grading of Sand

Louis Hanut, H. Bruyninckx

Robotic Intelligence

🎯 What it does: Researched and implemented an adaptive, iterative robot framework for sand profile grading aimed at achieving desired geometric curves

Robotic Mushroom Harvesting with Real2Sim2Real and Model Predictive Path Integral (MPPI) Based Planning

Konstantinos Vasios, Panagiotis Chatzakos

Domain AdaptationOptimizationRobotic IntelligenceAgriculture Related

🎯 What it does: Propose a robot button mushroom picking strategy based on the Real2Sim2Real pipeline and MPPI control planning, generating optimal root-pulling motion primitives.

Robotic Sim-to-Real Transfer for Long-Horizon Pick-and-Place Tasks in the Robotic Sim2Real Competition

Ming Yang, Yaran Chen

Domain AdaptationRobotic IntelligenceImage

🎯 What it does: A complete automated robotic system was developed to perform long-sequence pick-and-place tasks between simulation and real environments, including navigation, identification, grasping, and stacking, while maintaining consistent performance in multi-obstacle environments.

Robotic Space Simulator: Controls Implementation for Auxiliary Axes and Zero-G Dynamics

Eddie Hilburn, Robert Ambrose

Robotic IntelligencePhysics Related

🎯 What it does: Developed and tested two auxiliary axis control methods (Cartesian Workspace and Joint Cost-Function) as well as a mass property calculation and dynamic compensation method, improving the robot space simulator's handling of external torque inputs and zero-gravity behavior;

Robotic Tissue Manipulation in Endoscopic Submucosal Dissection Via Visual Feedback

Tao Zhang, Hamid Marvi

Robotic IntelligenceImage

🎯 What it does: Developed and evaluated a vision-based magnetic tissue manipulation system for endoscopic submucosal dissection (ESD), achieving precise tissue stretching and positioning through robotic arms equipped with small magnetic endoscopic clips attached to tissue and external large magnets in the ROS Gazebo simulation environment.

Robotic-CLIP: Fine-Tuning CLIP on Action Data for Robotic Applications

Nghia Nguyen, Anh Nguyen

Robotic IntelligenceVision Language ModelContrastive LearningVideoText

🎯 What it does: Fine-tune CLIP on a large amount of action video data to create Robotic-CLIP, enhancing the action understanding capability of robotic vision-language models.

Robots that Learn to Safely Influence via Prediction-Informed Reach-Avoid Dynamic Games

Ravi Pandya, Andrea V. Bajcsy

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed and solved a robust reach-avoid dynamic game, enabling robots to maximize their impact on humans while safety backup control is in place

Robots with Attitude: Singularity-Free Quaternion-Based Model-Predictive Control for Agile Legged Robots

Zixin Zhang, Zachary Manchester

OptimizationRobotic Intelligence

🎯 What it does: Proposed a singularity-free legged robot MPC framework using quaternions

Robust 4D Radar-Aided Inertial Navigation for Aerial Vehicles

Jinwen Zhu, Guoquan Huang

Autonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Developed a radar-inertial navigation system based on the error-state Kalman filter (ESKF), utilizing point-to-distribution radar scan matching and Doppler velocity measurements for tightly coupled state updates, achieving robust 4D millimeter-wave radar-assisted localization for drones.

Robust and Accurate Multi-View 2D/3D Image Registration with Differentiable X-Ray Rendering and Dual Cross-View Constraints

Yuxin Cui, Zhe Min

Pose EstimationImageBiomedical Data

🎯 What it does: Proposes a two-stage multi-view 2D/3D rigid registration method, where the first stage designs a joint loss function with cross-view constraints, and the second stage refines the pose through test-time optimization.

Robust Nonprehensile Dynamic Object Transportation: A Closed-Loop Sensitivity Approach

Ainoor Teimoorzadeh, Sami Haddadin

OptimizationRobotic Intelligence

🎯 What it does: Propose a closed-loop sensitivity optimization method to enhance the robustness of non-grasping dynamic transport tasks for robots, specifically for transporting objects that can freely move on tray-shaped end-effectors under partially known dynamic parameters.

Robust Optical Transceiver Manipulation in Cluttered Cable Environments Using 3D Scene Understanding and Planning

Iason Sarantopoulos, Antony I. T. Rowstron

Robotic IntelligenceImage

🎯 What it does: Designed a system for operating optical transceivers in cable-dense environments

Robust Orientation Control of Robot Manipulator Using Orientation Disturbance Observer

Kiyoung Choi, Sehoon Oh

Robotic Intelligence

🎯 What it does: A robust posture control algorithm for robot manipulators is proposed, achieving high-precision posture control through a disturbance observer (DOB) specifically designed for attitude dynamics.

Robust Planning for Autonomous Driving via Mixed Adversarial Diffusion Predictions

Albert Zhao, Stefano Soatto

Autonomous DrivingDiffusion model

🎯 What it does: Proposes a method for achieving strong robust planning in autonomous driving by hybridizing normal and adversarial predictions, which utilizes diffusion models to generate behavior predictions and evaluates plans using expected cost.

Robust Reinforcement Learning-Based Locomotion for Resource-Constrained Quadrupeds with Exteroceptive Sensing

Davide Plozza, Michele Magno

Robotic IntelligenceReinforcement LearningSimultaneous Localization and Mapping

🎯 What it does: Proposed a reinforcement learning-based external perception ground locomotion controller, utilizing real-time elevation mapping and depth sensors to enable stable walking of small quadruped robots on uneven terrain.

Robust Robot Walker: Learning Agile Locomotion over Tiny Traps

Shaoting Zhu, Hang Zhao

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: Proposed a method for quadruped robots to navigate various small obstacles (tiny traps) using only proprioceptive inputs, achieved through a two-stage training framework (comprising a contact encoder and a classification head), and designed a new benchmark test for this task.

Robust Robotic Breast Ultrasound Scanning and Real-Time Lesion Localization

Zhiyang Cao, Shaohua Zhang

Robotic IntelligenceBiomedical DataUltrasound

🎯 What it does: Developed a framework based on a finite state machine for autonomous breast ultrasound scanning, capable of dynamically switching between global scanning and detailed lesion scanning, employing radial and anti-radial scanning modes to achieve comprehensive coverage, while proposing a real-time detailed scanning method to avoid misjudgments caused by soft tissue motion, successfully detecting and locating lesions.

Robust Scene Change Detection Using Visual Foundation Models and Cross-Attention Mechanisms

Chun-Jung Lin, Feras Dayoub

SegmentationRepresentation LearningTransformerImage

🎯 What it does: Propose a scene change detection method based on the visual foundation model DINOv2 and the full-image cross-attention mechanism.

Robust Self-Reconfiguration for Fault-Tolerant Control of Modular Aerial Robot Systems

Rui Huang, Lin Zhao

OptimizationRobotic Intelligence

🎯 What it does: A robust and efficient self-reconfiguration algorithm for modular aerial robotic systems (MARS) is proposed, which maximizes the controllability margin in each intermediate phase and calculates the optimal disassembly/assembly sequence, verifying its effectiveness in various fault-tolerant self-reconfiguration scenarios.

Robust Swimming Controller for Soft Robots via Drop-Out Learning

Josephine Monica, Mark Campbell

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed a new framework enabling soft robotic fish to learn swimming even when actuators degrade or fail.

RoCoDA: Counterfactual Data Augmentation for Data-Efficient Robot Learning from Demonstrations

Ezra Ameperosa, Animesh Garg

Robotic Intelligence

🎯 What it does: Proposed and implemented a counterfactual data augmentation method called RoCoDA, integrating invariance, equivariance, and causality to enhance sample efficiency and generalization ability in robot imitation learning.

ROD: RGB-Only Fast and Efficient Off-Road Freespace Detection

Tong Sun, Yu Hu

SegmentationAutonomous DrivingTransformerImage

🎯 What it does: Proposed a novel method called ROD for detecting free space in off-road scenarios using only RGB images.

ROS2WASM: Bringing the Robot Operating System to the Web

Tobias Fischer, Michael Milford

Robotic Intelligence

🎯 What it does: Integrate RoboStack with WebAssembly to enable ROS 2 and its related software to run directly in the browser without requiring local installation, providing the www.ros2wasm.dev platform, support for the Robotics Toolbox for Python, and a browser-compatible version of the Swift simulator.

Routing Manipulation of Deformable Linear Object Using Reinforcement Learning and Diffusion Policy

Mingen Li, Changhyun Choi

Robotic IntelligenceReinforcement LearningDiffusion model

🎯 What it does: Propose a robust and fine-grained manipulation learning method for wrapping tasks involving deformable linear objects (DLOs), capable of completing the operation of threading a rope through a hole on rough surfaces and under unknown friction conditions.

RT-Affordance: Affordances are Versatile Intermediate Representations for Robot Manipulation

Soroush Nasiriany, Ted Xiao

Representation LearningRobotic IntelligenceSupervised Fine-TuningVision-Language-Action ModelImageTextMultimodality

🎯 What it does: Proposed a hierarchical model called RT-Affordance, which first generates joint pose planning (affordance plan) based on task language, then uses this plan as a conditional input to drive the robot to perform manipulation tasks. The model can combine heterogeneous supervision from large web datasets and robot trajectories, and learn new tasks using low-cost domain-specific affordance images without requiring additional collection of expensive robot trajectories.

RTAGrasp: Learning Task-Oriented Grasping from Human Videos via Retrieval, Transfer, and Alignment

Wenlong Dong, Hong Zhang

RetrievalDomain AdaptationRobotic IntelligenceTransformerVideoBenchmark

🎯 What it does: Proposes the RTAGrasp framework, which constructs robot memory using human grasp example videos and achieves training-free task-oriented grasping through three steps: retrieval, transfer, and alignment.

Run-time Observation Interventions Make Vision-Language-Action Models More Visually Robust

Asher Hancock, Anirudha Majumdar

RestorationVision-Language-Action ModelImageText

🎯 What it does: Designed and implemented a runtime intervention scheme called BYOVLA, which dynamically identifies sensitive regions in input images and uses an automatic image editing tool to minimize interference from task-irrelevant regions, thereby enhancing the robustness of Vision-Language-Action (VLA) models under visual disturbances.

S2BEV: Lightweight, Robust, and Precise SLAM-Oriented Segmentation Bird Eye's View Mapping Approach

Yefeng Sun, Chengliang Liu

Autonomous DrivingSimultaneous Localization and MappingAgriculture Related

🎯 What it does: Proposed a lightweight, robust, and accurate bird's-eye view segmentation mapping method called S2BEV, which integrates topological maps with semantic SLAM to construct orchard maps and support autonomous navigation.

Safe Control of Quadruped in Varying Dynamics via Safety Index Adaptation

Kai S. Yun, Changliu Liu

Robotic Intelligence

🎯 What it does: Proposed and implemented the Safe Index Adaptation (SIA) method, enabling quadruped robots to real-time update the safety index under varying loads, thereby achieving obstacle avoidance and performance objectives while ensuring the forward invariance of the safety region and finite-time convergence.

Safe Decentralized Multi-Agent Control using Black-Box Predictors, Conformal Decision Policies, and Control Barrier Functions

Sacha Huriot, Hussein Sibai

Safty and PrivacyRobotic IntelligenceVideo

🎯 What it does: Studied the use of uncertain black-box models to predict the trajectories of other agents in decentralized multi-agent robotic environments, and implemented safe control using safety constraints based on control barrier functions;

Safe Interval Motion Planning for Quadrotors in Dynamic Environments

Songhao Huang, Vijay Kumar

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Proposed a quadrotor safe interval motion planning framework for dynamic environments, generating trajectories through a two-stage process of front-end graph search and back-end gradient optimization.

Safe Leaf Manipulation for Accurate Shape and Pose Estimation of Occluded Fruits

Shaoxiong Yao, Kris Hauser

Pose EstimationRobotic IntelligenceAgriculture Related

🎯 What it does: Proposes an active fruit shape and pose estimation method, where the robot physically manipulates occluding leaves to expose hidden fruits and plans robot actions to maximize visibility while minimizing leaf damage; simultaneously develops scene-consistent shape completion techniques and perception-driven leaf deformation map models for planning.

Safe Multi-Agent Navigation Guided by Goal-Conditioned Safe Reinforcement Learning

Meng Feng, Brian C. Williams

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose a safe multi-agent navigation method combining planning and safe reinforcement learning

Safe Quadrotor Navigation Using Composite Control Barrier Functions

Marvin Harms, Kostas Alexis

OptimizationSafty and PrivacyRobotic IntelligencePoint Cloud

🎯 What it does: Proposes a safety filter based on Composite Control Barrier Functions (CCBF) to ensure safety in collision avoidance for multirotor drones, implemented on a third-order nonlinear dynamics model; analyzes the recursive feasibility of the safety filter under combined constraints and proves the infeasible set is negligible; demonstrates computational scalability under logarithmic constraints; and verifies through experiments that LiDAR-equipped quadrotors can maintain safe operation in crowded indoor and outdoor environments.

Safe Radial Segregation Algorithm for Swarms of Dubins-Like Robots

E. B. F. Filho, Luciano C. A. Pimenta

Robotic Intelligence

🎯 What it does: Designed and implemented a controller capable of radial separation for different types of Dubins-like robots;

SAFE-GIL: SAFEty Guided Imitation Learning for Robotic Systems

Yusuf Umut Ciftci, Somil Bansal

Adversarial AttackRobotic Intelligence

🎯 What it does: Propose SAFE-GIL, a design-time safety-guided behavior cloning method that injects adversarial perturbations during data collection to train safer control policies.

SafePCA: Enhancing Autonomous Robot Navigation in Dynamic Crowds Using Proximal Policy Optimization and Cellular Automata

A. Farouq, Joo-Ho Lee

Robotic IntelligenceReinforcement Learning

🎯 What it does: Developed the SafePCA framework, integrating cellular automata (CA) with a navigation method based on Proximal Policy Optimization (PPO), to enable safe and efficient robot navigation in dynamic crowds.

Safety and Naturalness Perceptions of Robot-to-Human Handovers Performed by Data-Driven Robotic Mimicry of Human Givers

Ava Megyeri, N. Banerjee

Safty and PrivacyRobotic Intelligence

🎯 What it does: Studies the safety and naturalness perception of robot-to-human (R2H) handover when robots imitate human actions in H2H handshake data.

Safety Guaranteed Robust Multi-Agent Reinforcement Learning with Hierarchical Control for Connected and Automated Vehicles

Zhili Zhang, Fei Miao

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose a safety-guaranteed hierarchical coordination and control scheme named Safe-RMM for collaborative control of connected autonomous vehicles in mixed traffic environments, considering system state uncertainty.

Safety-Critical Control for Aerial Physical Interaction in Uncertain Environment

Jeonghyun Byun, H. J. Kim

Robotic Intelligence

🎯 What it does: Designed a safety-critical control scheme based on a disturbance observer, using a safety filter to dynamically adjust the drone's attitude trajectory to achieve safe physical interaction.

Safety-Critical Control with Saliency Detection for Mobile Robots in Dynamic Multi-Obstacle Environments

Yu Zhang, A. Knoll

Robotic IntelligenceMultimodality

🎯 What it does: Propose a dual-filter architecture that utilizes RGB-D camera data and dynamic control potential functions (D-CBFs) to achieve real-time obstacle avoidance and safe control for mobile robots in dynamic environments with multiple obstacles;

Safety-Critical Locomotion of Biped Robots in Infeasible Paths: Overcoming Obstacles During Navigation Toward Destination

Jaemin Lee, A. Ames

Robotic Intelligence

🎯 What it does: Proposes a safety-critical walking control framework enabling legged robots to navigate safely from start to goal through infeasible paths in obstacle-dense environments.

Safety-Critical Online Quadrotor Trajectory Planner for Agile Flights in Unknown Environments

Jiazhe Yuan, Shuo Li

OptimizationRobotic Intelligence

🎯 What it does: A new method for achieving efficient, high-speed, collision-free quadrotor trajectory planning in unknown, crowded environments is proposed.

Safety-Critical Traffic Simulation with Adversarial Transfer of Driving Intentions

Zherui Huang, Xiao Sun

Autonomous DrivingOptimizationAdversarial AttackPoint Cloud

🎯 What it does: Propose the IntSim strategy, which decouples the driving intent of surrounding pedestrians or vehicles from their motion planning to achieve realistic and efficient simulation of safety-critical traffic scenarios; formalizes adversarial intent transfer as an optimization problem, enabling broad exploration of diverse attack behaviors; and employs powerful deep models with large-scale real data for intent-conditioned motion planning, combined with environment-adaptive intent adjustment, to realize dynamic adversarial interaction around autonomous vehicles; open-loop and closed-loop experiments conducted on real datasets such as nuScenes and Waymo demonstrate that IntSim achieves state-optimal simulation of safety-critical scenarios and further enhances the planner's performance in such scenarios.

SALON: Self-supervised Adaptive Learning for Off-road Navigation

Matthew Sivaprakasam, Sebastian A. Scherer

Domain AdaptationAutonomous Driving

🎯 What it does: Propose the SALON framework to achieve fast online adaptation for traversability estimation in off-road environments, minimizing human input and generating adaptive risk-aware cost and speed maps.

SAM-Guided Pseudo Label Enhancement for Multi-Modal 3D Semantic Segmentation

Mingyu Yang, Hun-Seok Kim

SegmentationDomain AdaptationMultimodality

🎯 What it does: Proposes a pseudo-label enhancement method based on SAM, leveraging 2D priors to improve cross-domain adaptation in multi-modal 3D semantic segmentation

Sample-Efficient Unsupervised Policy Cloning from Ensemble Self-Supervised Labeled Videos

Xin Liu, Yaran Chen

Reinforcement LearningVideoSequential

🎯 What it does: Propose the UPESV framework, which infers expert actions from unlabeled videos through a self-supervised video labeling model, subsequently clones strategies and collects environment interactions for self-supervised training, achieving sample-efficient unsupervised policy learning.

Sampling-Based Grasp and Collision Prediction for Assisted Teleoperation

Simon Manschitz, Dirk Ruiken

OptimizationRobotic Intelligence

🎯 What it does: Propose a learning-based shared autonomy framework that real-time tracks human-defined target poses and performs fine adjustments while satisfying dynamic constraints.

Sampling-Based Model Predictive Control for Volumetric Ablation in Robotic Laser Surgery

Vincent Y. Wang, L. Bridgeman

OptimizationRobotic Intelligence

🎯 What it does: Proposes a sampling-based model predictive control (MPC) scheme for planning laser ablation sequences for arbitrary tissue volumes.

SANDRO: A Robust Solver with a Splitting Strategy for Point Cloud Registration

Michael Adlerstein, Claudio Semini

Pose EstimationOptimizationPoint Cloud

🎯 What it does: Proposed a point cloud registration algorithm named SANDRO, combining the IRLS framework with a progressive non-convex robust loss function, and incorporating a splitting strategy to handle high outlier rates and skewed distributions of outliers.

SAP-SLAM: Semantic-Assisted Perception SLAM with 3D Gaussian Splatting

Yuheng Yang, Qingmin Liao

Gaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Proposed a dense SLAM system called SAP-SLAM that integrates high-fidelity reconstruction with advanced semantic understanding, utilizing 3D Gaussian Splatting technology to achieve differentiable rendering and scene representation;

SARO: Space-Aware Robot System for Terrain Crossing via Vision-Language Model

Shaoting Zhu, Hang Zhao

Robotic IntelligenceReinforcement LearningVision Language Model

🎯 What it does: Propose the SARO system, enabling quadruped robots to navigate from start to goal in 3D terrains through high-level reasoning and subtask closed-loop execution using vision-language models, while training low-level control strategies with probabilistic annealing selection methods.

SAS-Prompt: Large Language Models as Numerical Optimizers for Robot Self-Improvement

H. B. Amor (Arizona State University), Pannag R. Sanketi (Arizona State University)

OptimizationExplainability and InterpretabilityRobotic IntelligenceTransformerLarge Language ModelPrompt EngineeringRetrieval-Augmented Generation

🎯 What it does: Demonstrates the capability of large language models (LLMs) in iterative self-improvement of robot strategies, and proposes the SAS Prompt (Summarize, Analyze, Synthesize) method for generating new, unseen behaviors by retrieving, reasoning, and optimizing previous robot trajectories.

SaViD: Spectravista Aesthetic Vision Integration for Robust and Discerning 3D Object Detection in Challenging Environments

Tanmoy Dam, Mir Feroskhan

Object DetectionAutonomous DrivingImageMultimodalityPoint Cloud

🎯 What it does: Propose the SaViD framework, which employs a three-stage fusion alignment mechanism to enhance the robustness of LiDAR and camera fusion detection in long-range scenarios.

SayComply: Grounding Field Robotic Tasks in Operational Compliance Through Retrieval-Based Language Models

M. Ginting, A. Agha-mohammadi

Robotic IntelligenceTransformerRetrieval-Augmented Generation

🎯 What it does: Proposes SayComply, which employs retrieval-based language models to enforce operational compliance in robot task planning, and designs a hierarchical database and tree-structured retrieval-augmented generation (RAG) planner.