ICRA 2025 Papers — Page 12
IEEE International Conference on Robotics and Automation · 1604 papers
Pitching Motion in a Humanoid Robot Using Human-Inspired Shoulder Elastic Energy and Motor Torque Optimization
Yuri Nakazawa, A. Takanishi
OptimizationRobotic Intelligence
🎯 What it does: This study achieved a throwing motion similar to humans by adding elastic blades to the shoulder joint of a humanoid robot and optimizing motor torque;
PlanarNeRF: Online Learning of Planar Primitives with Neural Radiance Fields
Zheng Chen, Yi Xu
Object DetectionNeural Radiance Field
🎯 What it does: Propose the PlanarNeRF framework, which can detect dense 3D planes through online learning.
PlaneHEC: Efficient Hand-Eye Calibration for Multi-View Robotic Arm via Any Point Cloud Plane Detection
Ye Wang, Nanning Zheng
Computational EfficiencyRobotic IntelligencePoint Cloud
🎯 What it does: Proposed the PlaneHEC method, achieving efficient hand-eye calibration using only a depth camera and arbitrary planar surfaces
Planning for Tabletop Object Rearrangement
Jiaming Hu, Henrik I. Christensen
OptimizationRobotic Intelligence
🎯 What it does: This paper proposes an improved A*-based desktop object rearrangement algorithm that enhances the state representation and employs incremental goal attempts and lazy evaluation in each iteration to improve planning quality.
Planning with Adaptive World Models for Autonomous Driving
A. Vasudevan, Deva Ramanan
Autonomous DrivingGraph Neural NetworkWorld Model
🎯 What it does: Proposed BehaviorNet and AdaptiveDriver for implementing multi-agent adaptive planning in the nuPlan closed-loop simulator.
Planning with Linear Temporal Logic Specifications: Handling Quantifiable and Unquantifiable Uncertainty
Pian Yu, Marta Z. Kwiatkowska
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: The study addresses robot planning under quantifiable and non-quantifiable uncertainties, aiming to enable robots to optimally satisfy high-level tasks specified by linear temporal logic (LTL) formulas; it proposes a novel optimal robust strategy synthesis method for MDPSTs (Markov decision processes with set-valued transitions), introduces the concept of victory regions (WR), and provides an algorithm to compute WR on the product of MDPST and LDBA (limit-deterministic Büchi automata), followed by solving reachability problems using robust value iteration; the effectiveness and efficiency improvements of the method are validated through a case study of a mobile robot in a hexagonal world.
Planning-Oriented Cooperative Perception Among Heterogeneous Vehicles
Han Zheng, Yuanyuan Yang
Autonomous DrivingOptimizationBenchmark
🎯 What it does: Propose Scout, a planning-oriented early fusion collaborative perception framework for heterogeneous vehicles, and introduce Δθ-risk increment distribution (RID) and priority index (PI) to measure risk increments caused by incomplete perception, while developing a runtime estimation algorithm.
Plug-and-Play Multi-Domain Fusion Adaptation for Cross-Subject EEG-Based Motor Imagery Classification
Kecheng Shi, Jianwei Zhang
ClassificationDomain AdaptationGraph Neural NetworkSupervised Fine-TuningBiomedical Data
🎯 What it does: Proposed a plug-and-play multi-domain adaptation method that utilizes shared and private features to achieve rapid adaptation for cross-subject EEG motion imagination classification.
Plug-and-Play Physics-Informed Learning Using Uncertainty Quantified Port-Hamiltonian Models
Kaiyuan Tan, Thomas Beckers
Physics Related
🎯 What it does: A Plug-and-Play Physics-Informed Machine Learning (PnP-PIML) framework is proposed, which utilizes consistent prediction to detect anomalous dynamics and switches to distributed port-Hamiltonian systems (dPHS) at that time. The energy function is modeled via Gaussian processes, enabling the learning of system dynamics and Bayesian uncertainty quantification, thus providing reliable physics-consistent predictions in discrete distribution scenarios.
Pneumatic Logic Systems for Selectively Operating Distributed Pneumatic Elements
Rafael Ferrin Pozuelo, A. Kamimura
Physics Related
🎯 What it does: A novel standalone membrane valve was developed as a set/reset latch, and two pneumatic logic systems based on this valve were constructed, enabling the selection and parallel, independent operation of distributed pneumatic components with fewer pneumatic lines.
Point and Go: Intuitive Reference Frame Reallocation in Mode Switching for Assistive Robotics
Allie Wang, Martin Jägersand
Robotic Intelligence
🎯 What it does: Propose a mode switching scheme named Point and Go, which reassigns the Cartesian space reference frame into more intuitive translation and rotation modes, and defines a new translation axis through an innovative sweeping action pointing the gripper;
Point Cloud Decomposition for Task-Oriented Grasping
Khiem Phi, IV Ramakrishnan
Robotic IntelligencePoint Cloud
🎯 What it does: Proposes the ITSI (Iterative Slicing) method, which can segment multiple graspable regions within a single object point cloud without object knowledge, and achieve task-oriented grasping.
Point2Graph: An End-to-End Point Cloud-Based 3D Open-Vocabulary Scene Graph for Robot Navigation
Yifan Xu, Carol C. Menassa
ClassificationObject DetectionSegmentationRobotic IntelligencePoint Cloud
🎯 What it does: Proposes Point2Graph, an end-to-end 3D open-vocabulary scene graph generation framework based on point clouds, eliminating the dependency on pose-aligned RGB-D image sequences, and achieving room and object detection/segmentation as well as open-vocabulary classification through a hierarchical approach;
Points, Images and Texts: Boosting Point Cloud Completion with Multi-Modal Features
Chengkai Xia, Guang Chen
RestorationConvolutional Neural NetworkTransformerVision Language ModelImageTextMultimodalityPoint Cloud
🎯 What it does: Propose a point cloud completion method that leverages multimodal information from point clouds, images, and text. First, a visual question answering (VQA) model generates descriptive text for images, and a visual-text embedding model extracts joint features from image-text pairs. Meanwhile, multi-scale edge convolution is employed to describe geometric edge patterns, guiding local shape refinement. Cross-attention mechanisms are utilized to efficiently fuse multimodal features, further refining the coarse shape.
Points2Plans: From Point Clouds to Long-Horizon Plans with Composable Relational Dynamics
Yixuan Huang, Jeannette Bohg
Robotic IntelligenceTransformerLarge Language ModelWorld ModelTextPoint Cloud
🎯 What it does: Propose the Points2Plans framework, which uses a language model to generate high-level plans, employs a sampling planner to output continuous parameters that satisfy constraints, and achieves hierarchical planning for long-horizon manipulation tasks from partial viewpoint point clouds and language instructions.
PoLaRIS Dataset: A Maritime Object Detection and Tracking Dataset in Pohang Canal
Jiwon Choi, Younggun Cho
Object DetectionObject TrackingMultimodalityBenchmark
🎯 What it does: Proposed and constructed a new multimodal maritime obstacle detection and tracking dataset called PoLaRIS, and evaluated it using various state-of-the-art methods.
Polyhedral Collision Detection via Vertex Enumeration
Andrew Cinar, Forrest Laine
Optimization
🎯 What it does: Propose a framework for polyhedral collision detection, modeling the signed distance between two polyhedra as a convex optimization problem, and constructing high-level problem constraints by enumerating the vertices of related convex regions, avoiding the use of dedicated bi-level optimization solvers.
Polyp-Gen: Realistic and Diverse Polyp Image Generation for Endoscopic Dataset Expansion
Shengyuan Liu, Yixuan Yuan
GenerationData SynthesisDiffusion modelImageBiomedical Data
🎯 What it does: Propose an全自动 diffusion-based generation framework named Polyp-Gen to generate realistic and diverse polyp endoscopic images for dataset expansion.
PolyTouch: A Robust Multi-Modal Tactile Sensor for Contact-Rich Manipulation Using Tactile-Diffusion Policies
Jialiang Zhao, Edward H. Adelson
Robotic IntelligenceDiffusion modelMultimodality
🎯 What it does: Developed the PolyTouch multimodal tactile sensor and integrated it with vision and proprioception to construct a tactile diffusion strategy, enhancing robotic manipulation performance in contact-dense environments.
Portable, High-Frequency, and High-Voltage Control Circuits for Untethered Miniature Robots Driven by Dielectric Elastomer Actuators
Qi Shao, Huichan Zhao
Robotic IntelligenceVideo
🎯 What it does: A high-voltage high-frequency control circuit was proposed, using low-voltage resistors in series to control a 1.8kV DEA. Based on this circuit and a commercial micro high-voltage power converter, a 42g untethered crawling robot was developed, capable of crawling on desks and pipelines at a frequency of 15Hz while transmitting video in real-time.
Potential Fields as Scene Affordance for Behavior Change-Based Visual Risk Object Identification
Pang-Yuan Pao, Yi-Ting Chen
Object DetectionSegmentationAutonomous DrivingImageBenchmark
🎯 What it does: Explored visual risk object recognition based on behavior changes and proposed a novel recognition framework leveraging bird's-eye view and potential field models.
Pre-Surgical Planner for Robot-Assisted Vitreoretinal Surgery: Integrating Eye Posture, Robot Position and Insertion Point
Satoshi Inagaki, M. Nasseri
OptimizationRobotic Intelligence
🎯 What it does: Designed and verified an optimization framework for eye tilt and robot positioning in robot-assisted vitreoretinal surgery.
Precision Harvesting in Cluttered Environments: Integrating End Effector Design with Dual Camera Perception
Kendall Koe, Girish Chowdhary
Robotic IntelligenceImageAgriculture Related
🎯 What it does: Propose a collaborative design framework for precise harvesting in cluttered environments, integrating a global detection camera and a local eye-in-hand camera to achieve closed-loop visual feedback and reliable error handling for small fruits.
Predictive Kinematic Coordinate Control for Aerial Manipulators Based on Modified Kinematics Learning
Zhengzhen Li, Shiyu Zhao
Robotic Intelligence
🎯 What it does: This paper proposes a predictive kinematic coordinate control method for aerial manipulators, which includes a learning-based improved kinematic model and a model predictive control (MPC) scheme based on weight allocation.
Prepared for the Worst: Resilience Analysis of the ICP Algorithm via Learning-Based Worst-Case Adversarial Attacks
Ziyu Zhang, Timothy D. Barfoot
Adversarial AttackPoint Cloud
🎯 What it does: Proposed a learning-based worst-case attack method to evaluate the robustness of LiDAR point cloud ICP algorithms.
PRESTO: Fast Motion Planning Using Diffusion Models Based on Key-Configuration Environment Representation
Mingyo Seo, Beomjoon Kim
OptimizationRobotic IntelligenceDiffusion model
🎯 What it does: Propose a learning-guided motion planning framework that generates seed trajectories using diffusion models and refines them through trajectory optimization, approximating C-space obstacles with sparse key configurations.
PRIDEV: A Plug-and-Play Refinement for Improved Depth Estimation in Videos
Jing Xu, Xinhua Xu
Depth EstimationVideo
🎯 What it does: Propose a pluggable depth estimation refinement method PRIDEV, which leverages the robustness of image depth estimation models to seamlessly transfer to video depth estimation, and introduces a Time Depth Stabilization Module (TDSM) to capture temporal features in videos.
PRIMER: Perception-Aware Robust Learning-Based Multiagent Trajectory Planner
Kota Kondo, Jonathan P. How
Autonomous DrivingOptimizationComputational Efficiency
🎯 What it does: Proposed a perception-aware distributed asynchronous multi-agent trajectory planning method called PARM and PARM*, and introduced PRIMER based on imitation learning, enabling efficient trajectory planning in uncertain environments;
Privileged-Dreamer: Explicit Imagination of Privileged Information for Rapid Adaptation of Learned Policies
Morgan Byrd, Sehoon Ha
Recurrent Neural NetworkReinforcement Learning
🎯 What it does: Propose PrivilegedDreamer, a model-driven reinforcement learning framework that explicitly estimates hidden parameters and utilizes them in the model, actor, and critic networks.
Proactive Tactile Exploration for Object-Agnostic Shape Reconstruction from Minimal Visual Priors
Paris Oikonomou, C. Tzafestas
GenerationRobotic IntelligenceMesh
🎯 What it does: Proposes an iterative method for 3D shape reconstruction based on active tactile exploration, first fitting data points using a single geometric template and then locally adjusting the mesh to capture deformations.
Probabilistic Latent Variable Modeling for Dynamic Friction Identification and Estimation
V. Vantilborgh, G. Crevecoeur
Robotic Intelligence
🎯 What it does: Proposed and implemented a probabilistic latent variable model based on latent dynamic states for dynamic friction identification and estimation in robot joints.
PROBE: Proprioceptive Obstacle Detection and Estimation while Navigating in Clutter
Dhruv Metha Ramesh, Abdeslam Boularias
Robotic IntelligenceTransformerTime SeriesSequential
🎯 What it does: Propose a technology called PROBE that utilizes only the robot's own perception to detect and estimate occluded rectangular obstacles along with their size and pose in crowded environments.
ProDapt: Proprioceptive Adaptation Using Long-Term Memory Diffusion
Federico Pizarro Bejarano, Angela P. Schoellig
Robotic IntelligenceDiffusion model
🎯 What it does: Proposed the ProDapt method, which introduces long-term memory of past interactions between the robot and the environment into diffusion models, enabling task completion based solely on proprioceptive sensing
Promi: an Efficient Prototype-Mixture Baseline for Few-Shot Segmentation with Bounding-Box Annotations
Florent Chiaroni, Ola Ahmad
SegmentationMixture of ExpertsImage
🎯 What it does: Propose a few-shot binary segmentation method ProMi based on bounding box annotations, utilizing a prototype-mixture model to handle background classes
Prompt-Responsive Object Retrieval with Memory-Augmented Student-Teacher Learning
Malte Mosbach, Sven Behnke
RetrievalRobotic IntelligenceRecurrent Neural NetworkReinforcement LearningPrompt EngineeringImage
🎯 What it does: Propose a framework that combines a prompt-manipulable foundation model with reinforcement learning, enabling robots to perform fine-grained manipulation tasks based on user prompts.
Propagative Distance Optimization for Motion Planning
Yu Chen, Guanya Shi
OptimizationRobotic Intelligence
🎯 What it does: Proposes the PDOMP method, which utilizes distance-based kinematic expressions and chain structures to achieve more efficient forward kinematics and Jacobian matrix calculations in serial joint robot motion planning.
Proprioceptive Object Shape and Size Extraction via In-Hand-Manipulation with a Variable Friction Robot Gripper
I. Bodnar, Adam Spiers
Robotic Intelligence
🎯 What it does: Using a 3-DOF variable friction gripper, shape and size extraction of 2D convex polyhedra is achieved through in-hand manipulation and proprioception
Provable Methods for Searching with an Imperfect Sensor
Nilanjan Chakraborty, Michael Perk
OptimizationRobotic Intelligence
🎯 What it does: This paper studies the problem of finding a target within a finite set of positions on a plane using a mobile robot equipped with imperfect sensors under a finite time budget, and proposes a fast algorithm along with corresponding performance guarantees; meanwhile, it provides complexity analysis for the problem and its variants and conducts experimental evaluation.
ProxFly: Robust Control for Close Proximity Quadcopter Flight Via Residual Reinforcement Learning
Ruiqi Zhang, Mark W. Mueller
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed the ProxFly residual deep reinforcement learning controller to compensate for downwash effects and enhance quadrotor flight stability during close-range flights.
Proximity and Visuotactile Point Cloud Fusion for Contact Patches in Extreme Deformation
Jessica Yin, Russ Tedrake
SegmentationMultimodalityPoint Cloud
🎯 What it does: Proposed a proximity sensing and visual-tactile sensing point cloud fusion algorithm for contact patch segmentation, completely independent of membrane mechanics.
PseudoTouch: Efficiently Imaging the Surface Feel of Objects for Robotic Manipulation
Adrian Röfer, A. Valada
RecognitionRepresentation LearningRobotic IntelligenceSupervised Fine-TuningImageMultimodality
🎯 What it does: Designed and trained the PseudoTouch model to rapidly infer tactile signals from depth images through low-dimensional visual-tactile embeddings, and evaluated it on object recognition and grasp stability prediction tasks.
PTQ4RIS: Post-Training Quantization for Referring Image Segmentation
Xiaoyan Jiang, Sifan Zhou
SegmentationVision Language ModelMultimodality
🎯 What it does: This paper proposes a post-training quantization framework called PTQ4RIS specifically for referential image segmentation tasks and analyzes the root causes of performance degradation during the quantization process.
PTZ-Calib: Robust Pan-Tilt-Zoom Camera Calibration
Jinhui Guo, Jieping Ye
Pose EstimationComputational Efficiency
🎯 What it does: Proposes a two-stage PTZ camera calibration method called PTZ-Calib, which can efficiently and accurately estimate camera parameters from any perspective.
PUGS: Perceptual Uncertainty for Grasp Selection in Underwater Environments
Onur Bagoren, Aaron Marburg
Depth EstimationRobotic Intelligence
🎯 What it does: Propose a framework that quantifies and represents perceptual uncertainty in 3D reconstruction through occupancy uncertainty estimation, integrating it into autonomous grasping selection in underwater environments.
PUGS: Zero-Shot Physical Understanding with Gaussian Splatting
Yinghao Shuai, Hao Zhao
Representation LearningContrastive LearningGaussian SplattingBenchmarkPhysics Related
🎯 What it does: Reconstruct 3D objects using Gaussian splatting and predict their physical properties such as mass and friction in a zero-shot scenario.
Pushing Through Clutter with Movability Awareness of Blocking Obstacles
Joris J. Weeda, Javier Alonso-Mora
OptimizationRobotic Intelligence
🎯 What it does: Propose a mobility-aware planning framework combining global semantic visual graph and local model predictive path integral (SVG-MPPI) for solving navigation problems with movable obstacles.
Qdgset: a Large Scale Grasping Dataset Generated With Quality-Diversity
J. Huber, Stéphane Doncieux
Data SynthesisRobotic IntelligenceMesh
🎯 What it does: Extend the QDG-6DoF quality diversity (QD) framework by integrating data augmentation methods combining mesh transformation and transfer learning to generate a large-scale 6DoF grasping dataset called QDGset.
QuadWBG: Generalizable Quadrupedal Whole-Body Grasping
Jilong Wang, He Wang
Robotic IntelligenceReinforcement LearningImage
🎯 What it does: Propose a modular framework based on a single-arm camera to achieve overall body movement and grasping control in quadruped robots.
Quarry-Bot: A Reconfigurable Cable-Suspended Robot for Lunar Site Engineering
Zahir Abram Castrejon, Paul Y. Oh
OptimizationRobotic Intelligence
🎯 What it does: Developed a reconfigurable rope-hanging robot called Quarry-Bot capable of autonomously clearing debris on the lunar surface, preparing for future base and infrastructure construction.
Quart-Online: Latency-Free Multimodal Large Language Model for Quadruped Robot Learning
Xinyang Tong, Shangke Lyu
Computational EfficiencyRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningVision-Language-Action ModelImageTextMultimodality
🎯 What it does: Proposed a latency-free quadruped robot multimodal large language model named QUART-Online to enhance reasoning efficiency in visual-language-action tasks for quadruped robots.
QuasiNav: Asymmetric Cost-Aware Navigation Planning with Constrained Quasimetric Reinforcement Learning
Jumman Hossain, Nirmalya Roy
Autonomous DrivingReinforcement Learning
🎯 What it does: This paper proposes the QuasiNav framework, which achieves anisotropic cost-aware navigation planning using quasi-metric embedding and constrained reinforcement learning.
QueryCAD: Grounded Question Answering for CAD Models
Claudius Kienle, Rainer Jäkel
SegmentationRobotic IntelligenceVision Language ModelMeshBenchmark
🎯 What it does: Proposed the QueryCAD system, which enables precise information extraction from CAD models through natural language queries; the system integrates the SegCAD open-vocabulary instance segmentation model, capable of identifying and selecting specific components in the model based on component descriptions; meanwhile, a CAD question-answering benchmark was designed, and QueryCAD was embedded into an automatic robot program synthesis framework to verify its ability to enhance deep learning-based robotic solutions.
QuickGrasp: Lightweight Antipodal Grasp Planning with Point Clouds
Navin Sriram Ravie, Bijo Sebastian
SegmentationOptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Propose a lightweight analytical method for anti-symmetric grasping planning, primarily by optimizing to find grasp points on object surfaces, combined with a soft region growth algorithm for planar segmentation, and using an optimization-based quality metric to evaluate grasp points to ensure indirect force closure.
QVIO2: Quantized Map-Based Visual-Inertial Odometry
Yuxiang Peng, Guoquan Huang
Autonomous DrivingOptimizationSimultaneous Localization and Mapping
🎯 What it does: Proposed QVIO2, a quantized map visual-inertial odometry system that improves data quantization strategies and uniformly processes multi-bit and single-bit, original and residual quantized measurements.
R+X: Retrieval and Execution from Everyday Human Videos
Georgios Papagiannis, Edward Johns
RetrievalRobotic IntelligenceVision Language ModelVideoText
🎯 What it does: The R+X framework enables robots to learn skills from long-duration unannotated first-person human daily task videos; upon receiving a language instruction, it first retrieves relevant short video clips and then executes the skill through context imitation learning.
RACCOON: Grounding Embodied Question-Answering with State Summaries from Existing Robot Modules
Samuel Bustamante, F. Stulp
Robotic IntelligenceLarge Language ModelVision Language ModelRetrieval-Augmented Generation
🎯 What it does: Propose a framework named RACCOON that combines answers from a base model with the robot's internal knowledge to more accurately answer arbitrary queries about the environment and interactions.
RaccoonBot: An Autonomous Wire-Traversing Solar-Tracking Robot for Persistent Environmental Monitoring
Efrain Mendez-Flores, Magnus Egerstedt
Robotic Intelligence
🎯 What it does: Proposed RaccoonBot, an autonomous solar tracking robot that walks along wires for continuous environmental monitoring.
RACE: A Fast and Lightweight Urban Exploration and Search Strategy for Multi-Robot Systems
J. L. Kit (Singapore University of Technology and Design), G. Soh (Singapore University of Technology and Design)
OptimizationRobotic Intelligence
🎯 What it does: Developed a lightweight multi-robot urban exploration and search algorithm RACE based on improved ant colony optimization, which can efficiently perform collaborative exploration in unknown indoor environments.
RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning
Yinpei Dai, Joyce Chai
Robotic IntelligenceVision Language Model
🎯 What it does: Propose a scalable data generation pipeline that automatically expands expert demonstrations into failure recovery trajectories and fine-grained language annotations, and introduce the RACER framework, which consists of an online visual-language model supervisor and a language-conditioned visual-motor policy, to enhance the robustness and error-correctability of robotic visual-motor policies.
Radar Teach and Repeat: Architecture and Initial Field Testing
Xinyuan Qiao, Tim D. Barfoot
Autonomous DrivingRobotic IntelligenceSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Designed and verified a full-stack radar system, Radar Teach and Repeat (RT&R), achieving GPS-free offline long-distance autonomous robot path tracking;
Radar4VoxMap: Accurate Odometry from Blurred Radar Observations
Ji-Kwon Seok, Kichun Jo
Autonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposed the Radar4VoxMap method to enhance the accuracy of odometry using only radar.
RadarMask: A Novel End-to-End Sparse Millimeter-Wave Radar Sequence Panoptic Segmentation and Tracking Method
Yubo Guo, Qiang Gao
Object TrackingSegmentationPoint Cloud
🎯 What it does: Proposed an end-to-end method for panoptic segmentation and tracking of sparse millimeter-wave radar sequences, RadarMask, which can achieve multi-level semantic and instance descriptions in the radar data domain;
RaggeDi: Diffusion-Based State Estimation of Disordered Rags, Sheets, Towels and Blankets
Jikai Ye, G. Chirikjian
Image TranslationGenerationDiffusion modelImage
🎯 What it does: This paper proposes a diffusion model-based approach that transforms fabric state estimation into an image generation problem. It uses RGB images to represent point-to-point translations (translation maps) from a predefined flat mesh to a deformed mesh, and trains a conditional diffusion model to predict this translation map based on observations.
RAIL: Reachability-Aided Imitation Learning for Safe Policy Execution
Wonsuhk Jung, Shreyas Kousik
Safty and PrivacyRobotic Intelligence
🎯 What it does: Proposes the RAIL method for safe execution of IL based on reachability safety filters, aiming to enforce hard constraints and ensure robot safety during task execution.
Range-Based 6-DoF Monte Carlo SLAM with Gradient-Guided Particle Filter on GPU
Takumi Nakao, Hisashi Date
Computational EfficiencySimultaneous Localization and Mapping
🎯 What it does: Proposed a gradient-guided particle filter update strategy and keyframe map representation for achieving high-dimensional 6-DoF Monte Carlo SLAM.
Rao-Blackwellized POMDP Planning
Jiho Lee, Zachary Sunberg
Autonomous DrivingReinforcement Learning
🎯 What it does: This paper proposes an approximate solver for Rao-Blackwellized POMDP (RB-POMDP), applying Rao-Blackwellization techniques in Bayesian updates and online planning, and subsequently compares it with the traditional SIRPF in a simulated GPS failure localization task;
Rapid Autonomous Exploration of Large-Scale Environments for Ground Robots Based on Region Partitioning
Zhi Wen, Jing Liu
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposed a hierarchical planning and region partitioning based autonomous exploration method for ground robots.
Rapid Dynamic Obstacle Avoidance for UAVs Enhanced by DVS and Neuromorphic Computing
Siyang Wang, Pengju Ren
Computational EfficiencyRobotic IntelligenceSpiking Neural Network
🎯 What it does: Developed an end-to-end obstacle avoidance algorithm for UAVs based on monocular dynamic visual sensors
Rapid Online Learning of Hip Exoskeleton Assistance Preferences
Giulia Ramella, Mohamed Bouri
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningBiomedical Data
🎯 What it does: Quickly learn the assistive preferences of knee exoskeletons by performing pairwise comparisons of randomly generated assistive configurations and actively querying user preferences.
RAR-6: An Optimized Reconfigurable Asymmetric 6-DOF Haptic Robot for Gross and Fine Motor Tasks
Changqi Zhang, Mingming Zhang
OptimizationRobotic Intelligence
🎯 What it does: Designed and validated a 6-DOF tactile robot based on a reconfigurable asymmetric parallel mechanism for gross motor tasks and fine motor tasks.
RE-TRIP: Reflectivity Instance Augmented Triangle Descriptor for 3D Place Recognition
Yechan Park, Euntai Kim
RetrievalPoint Cloud
🎯 What it does: Proposed a 3D position recognition descriptor RE-TRIP that integrates reflectance and geometric information, achieving keypoint extraction, instance segmentation, matching, and reflectance-based loop closure verification.
RE0: Recognize Everything with 3D Zero-Shot Instance Segmentation
Xiaohan Yan, Zhicheng Wang
SegmentationTransformerVision Language ModelImageMultimodalityPoint Cloud
🎯 What it does: Proposes a 3D zero-shot instance segmentation framework named RE0, which utilizes multi-view RGB-D images, point cloud geometric information, projection relationships, and CLIP semantic features to achieve label assignment from 2D to 3D and instance consistency;
Reachability Analysis for Black-Box Dynamical Systems
Vamsi Krishna Chilakamarri, Somil Bansal
OptimizationSafty and Privacy
🎯 What it does: Propose a reachability analysis method for black-box dynamical systems, which approximates the Hamiltonian function using samples and solves the Hamilton-Jacobi partial differential equation to obtain the reachable set and safe controller.
REACT: Multi Robot Energy-Aware Orchestrator for Indoor Search and Rescue Critical Tasks
Fabio Maresca, Xavier Pérez Costa
OptimizationRobotic Intelligence
🎯 What it does: Propose the REACT energy-aware orchestrator to optimize the exploration phase in search and rescue tasks, extending robot operational time and improving area coverage.
Reactive Collision Avoidance for Safe Agile Navigation
Alessandro Saviolo, Giuseppe Loianno
Autonomous DrivingOptimizationImagePoint Cloud
🎯 What it does: Proposes a unified real-time collision avoidance framework that integrates perception, planning, and control into a single system using onboard sensing and computing; the framework employs neural networks to refine noisy RGB-D data, extracts the minimum collision time point as a constraint, and dynamically adjusts optimization through heuristic methods to maintain a balance between safety and agility.
Real-Time 3D Reconstruction via Camera-Lidar (2D) Fusion for Mobile Robots: A Gaussian Splatting Approach
Ajay Kumar Sandula, Pradipta Biswas
Autonomous DrivingRobotic IntelligenceGaussian SplattingSimultaneous Localization and MappingImageMultimodality
🎯 What it does: Proposed a real-time 3D reconstruction SLAM method that integrates camera and 2D LiDAR fusion with Gaussian Splatting technology
Real-time Deformation-aware Control for Autonomous Robotic Subretinal Injection under iOCT Guidance
Demir Arikan, I. Iordachita
Robotic IntelligenceBiomedical Data
🎯 What it does: Proposed an autonomous subretinal injection method under iOCT guidance with real-time deformation awareness.
Real-Time Grasp Quality in Boundary-Constrained Granular Swarm Robots
Declan Mulroy, Ankit Srivastava
Robotic Intelligence
🎯 What it does: Designed a soft robot for boundary-constrained granular swarm robots, achieving motion and grasping functionalities.
Real-Time LiDAR Point Cloud Compression and Transmission for Resource-Constrained Robots
Yuhao Cao, Haoyao Chen
CompressionPoint Cloud
🎯 What it does: Proposed a real-time LiDAR point cloud compression and transmission framework named RCPCC.
Real-Time Safe Bipedal Robot Navigation using Linear Discrete Control Barrier Functions
Chengyang Peng, Ayonga Hereid
Robotic Intelligence
🎯 What it does: Developed a unified safe path and gait planning framework capable of online real-time evaluation
Real-Time Sampling-based Online Planning for Drone Interception
Gilhyun Ryou, S. Karaman
Autonomous DrivingOptimization
🎯 What it does: Propose a sampling-based online planning algorithm that uses neural network inference instead of time-consuming nonlinear trajectory optimization to quickly generate multiple trajectories for different possible positions of the target drone, and selects the shortest reachable trajectory for interception by comparing with the predicted arrival time of the target.
Real-Time UAV Tracking: A Comparative Study of YOLOv8 with Object Tracking Algorithms
T. Russo, Nikolaos Vitzilaios
Object TrackingComputational EfficiencyConvolutional Neural NetworkVideoBenchmark
🎯 What it does: This paper conducts experimental comparisons of YOLOv8 combined with various target tracking algorithms for drone tracking, evaluates their performance; simultaneously optimizes model size to balance speed and accuracy; and implements a real-time tracking system on the Jetson Orion Nano GPU;
Real-Time Whole-Body Control of Legged Robots with Model-Predictive Path Integral Control
Juan Alvarez-Padilla, Zachary Manchester
Robotic Intelligence
🎯 What it does: Proposes a system capable of synthesizing whole-body gaits and manipulation strategies for real-world multi-legged robots in real-time conditions, achieving full-body motion and manipulation through sampling-based model predictive path integral control (MPI);
Real-World Automated Vehicle Longitudinal Stability Analysis: Controller Design and Field Test
Ke Ma, Xiaopeng Li
Autonomous Driving
🎯 What it does: On a real-world autonomous vehicle platform, traditional delay-free following controllers are compared with longitudinal controllers that consider vehicle dynamic responses. The stability conditions of both are calculated, and their performance in suppressing traffic oscillations is validated through field experiments.
Realistic Extreme Behavior Generation for Improved AV Testing
Robert Dyro, Marco Pavone
GenerationAutonomous DrivingAdversarial Attack
🎯 What it does: Proposes a framework that leverages real-world collision-free data to generate synthetic extreme collision scenarios for evaluating autonomous vehicle (AV) collision avoidance technologies
Realizing Emergent Collective Behaviors Through Robotic Swarmalators
R. Beattie, Daniela Rus
Robotic Intelligence
🎯 What it does: Implemented various Swarmalator behaviors using a 15-robot experimental platform, including rotational and non-rotational, frequency coupling, and uniform vs. non-uniform natural frequency distributions.
Realm: Real-Time Line-of-Sight Maintenance in Multi-Robot Navigation with Unknown Obstacles
Ruofei Bai, Lihua Xie
Robotic IntelligencePoint Cloud
🎯 What it does: Proposes a multi-robot navigation method based on real-time LiDAR scans with Line-of-Sight (LoS) constraints, capable of deriving LoS constraints in real-time through point cloud visibility analysis in unknown environments;
Realtime Limb Trajectory Optimization for Humanoid Running Through Centroidal Angular Momentum Dynamics
Sait Sovukluk, Christian Ott
OptimizationRobotic Intelligence
🎯 What it does: Studied and proposed a real-time nonlinear optimization problem for quadrupedal locomotion in humanoid robots, and tested and validated the optimized trajectories on two different humanoid robot models.
Reduced-Order Model Guided Contact-Implicit Model Predictive Control for Humanoid Locomotion
Sergio A. Esteban, Aaron D. Ames
OptimizationRobotic Intelligence
🎯 What it does: Proposes a control framework that combines the Hybrid Linear Inverted Pendulum (HLIP) with Contact-Implicit Model Predictive Control (CI-MPC) for gait generation and control of the 24-degree-of-freedom humanoid robot Achilles; the framework's robustness is validated through simulation under uneven terrain, disturbance recovery, model and state uncertainty, and interaction with environmental obstacles, while operating in real-time at 50Hz.
Reduced-Order Model-Based Gait Generation for Snake Robot Locomotion Using NMPC
Adarsh Salagame, Alireza Ramezani
OptimizationRobotic Intelligence
🎯 What it does: Proposes an optimization-based motion planning method designed for snake robots operating in confined environments, utilizing a simplified reduced-order model to streamline the planning process, enabling the optimizer to autonomously generate gaits while controlling the robot's footprint within narrow spaces;
Reference-Free Formula Drift with Reinforcement Learning: From Driving Data to Tire Energy-Inspired, Real-World Policies
Franck Djeumou, John K. Subosits
Autonomous DrivingReinforcement LearningStochastic Differential Equation
🎯 What it does: Designed and trained a reinforcement learning agent that utilizes the tire energy absorption concept to achieve real-time drifting, maintaining the vehicle within the track boundaries in complex and varying waypoint configurations; and achieved zero-shot deployment on real vehicles through training in a simulation environment based on a neural stochastic differential equations (SDEs) vehicle model.
Reinforcement Learning Driven Multi-Robot Exploration via Explicit Communication and Density-Based Frontier Search
Gabriele Calzolari, G. Nikolakopoulos
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposing a decentralized collaborative multi-robot exploration framework based on reinforcement learning
Reinforcement Learning for Adaptive Planner Parameter Tuning: A Perspective on Hierarchical Architecture
Wangtao Lu, Yue Wang
OptimizationReinforcement Learning
🎯 What it does: Proposed a hierarchical architecture based on reinforcement learning for automatically tuning planner parameters.
Reinforcement Learning on Reconfigurable Hardware: Overcoming Material Variability in Laser Material Processing
Giulio Masinelli, David Atienza
OptimizationReinforcement LearningPhysics Related
🎯 What it does: Propose a real-time reinforcement learning method implemented on FPGA for laser processing control to adapt to changes in material and surface properties.
Reinforcement Learning with Lie Group Orientations for Robotics
Martin Schuck, Angela P. Schoellig
Representation LearningRobotic IntelligenceReinforcement Learning
🎯 What it does: This paper studies the use of Lie group structures to handle the orientation problems of robots and objects, and proposes simple modifications to the network's input and output within a reinforcement learning framework to align with the Lie group structure of orientation, achieving an easily implementable, compatible with existing learning libraries, and significantly improved performance solution.
Reinforcement Learning Within the Classical Robotics Stack: A Case Study in Robot Soccer
Adam Labiosa, Josiah P. Hanna
Robotic IntelligenceReinforcement Learning
🎯 What it does: Developed a new architecture that integrates model-agnostic reinforcement learning (RL) into the classic robot stack, employing a multi-fidelity sim2real approach and decomposing behavior into learned sub-behaviors combined with heuristic selection;
RelAIBotiX: Reliability Assessment for AI-Controlled Robotic Systems
Philipp Grimmeisen, Andrey Morozov
Robotic Intelligence
🎯 What it does: A new method named RelAIBotiX is proposed for dynamically and continuously evaluating the reliability of AI-controlled robotic systems, comprising four sub-methods: skill detection, behavior analysis, reliability model generation, and model solving.
Relative Velocity-Based Reward Model for Socially-Aware Navigation with Deep Reinforcement Learning
Vinu Maddumage, Jodi Martin
Reinforcement Learning
🎯 What it does: Propose a reward model based on relative velocity for collision avoidance when mobile robots coexist with humans in shared environments, and compare it with traditional reward models in simulation environments.
Relevance-Driven Decision Making for Safer and More Efficient Human Robot Collaboration
Xiaotong Zhang, Kamal Youcef-Toumi
Safty and PrivacyRobotic IntelligenceTransformerLarge Language Model
🎯 What it does: Propose a decision-making method based on relevance and develop a dual-loop framework (real-time and asynchronous) for scene understanding, human intent prediction, and safe and efficient decision-making in human-robot collaboration.
Reliable Aerial Manipulation: Combining Visual Tracking with Range Sensing for Robust Grasping
Marc Blöchlinger, Robert K. Katzschmann
Object TrackingDepth EstimationRobotic IntelligencePoint Cloud
🎯 What it does: Combined 1D time-of-flight (TOF) distance sensor with visual positioning system for UAV aerial grasping tasks, achieving reliable object localization