arXivSub Start free trial

ICRA 2025 Papers — Page 11

IEEE International Conference on Robotics and Automation · 1604 papers

Neural Lyapunov Function Approximation with Self-Supervised Reinforcement Learning

Lucy McCutcheon, Saber Fallah

OptimizationRobotic IntelligenceReinforcement LearningWorld ModelSequential

🎯 What it does: Proposes a method for neural network approximation of nonlinear Lyapunov functions using self-supervised reinforcement learning and data-driven world models.

Neural Ranging Inertial Odometry

Si Wang, Yue Wang

Recurrent Neural NetworkGraph Neural NetworkSimultaneous Localization and Mapping

🎯 What it does: Proposed a new neural fusion framework for ranging inertial odometry, combining graph attention UWB network and recurrent neural inertial network

Neural Surface Reconstruction and Rendering for LiDAR-Visual Systems

Jianheng Liu, Fu Zhang

GenerationNeural Radiance FieldImagePoint Cloud

🎯 What it does: Propose a unified surface reconstruction and rendering framework that integrates NeRF and NDF to recover appearance and structural information from pose images and point clouds.

Neuraloc: Visual Localization in Neural Implicit Map With Dual Complementary Features

Hongjia Zhai, Guofeng Zhang

Pose EstimationNeural Radiance FieldSimultaneous Localization and MappingImage

🎯 What it does: Proposes a visual localization method based on neural implicit maps, utilizing dual complementary features to achieve efficient 2D-3D correspondence matching, and enhancing geometric constraints and semantic information through 3D keypoint descriptor fields and semantic context feature fields.

Neuro-Symbolic Imitation Learning: Discovering Symbolic Abstractions for Skill Learning

Leon Keller, Jan Peters

Robotic IntelligenceReinforcement Learning from Human Feedback

🎯 What it does: Propose a neuro-symbolic imitation learning framework that first learns symbolic abstract representations through task demonstrations, then generates abstract plans using symbolic planning, and finally refines the abstract plans into executable robot commands via neural skills.

New Graph Distance Measures and Matching of Topological Maps for Robotic Exploration

Fabio Morbidi

Robotic IntelligenceSimultaneous Localization and MappingGraph

🎯 What it does: Proposed three new graph distance metrics that satisfy metric requirements and applied them to topological graph matching in robot exploration

Next Best Sense: Guiding Vision and Touch with FisherRF for 3D Gaussian Splatting

M. Strong, Monroe Kennedy

Robotic IntelligenceGaussian Splatting

🎯 What it does: Proposed an active next best view and contact selection framework based on 3D Gaussian Splatting (3DGS) for robotic manipulators.

Next-Best-Trajectory Planning of Robot Manipulators for Effective Observation and Exploration

Heiko Renz, Torsten Bertram

Robotic Intelligence

🎯 What it does: Developed a next-best-trajectory planning strategy for robot arms in dynamic environments, generating local trajectories to maximize observational information and avoid collisions, combined with voxel maps and multi-view ray casting.

Nezha-Mb: Design and Implementation of a Morphing Hybrid Aerial-Underwater Vehicle

Zhuxiu Xu, Zheng Zeng

Robotic Intelligence

🎯 What it does: Designed and implemented a deformable hybrid aerial-underwater robot named Nezha-MB capable of smoothly transitioning between air and water.

No Plan but Everything Under Control: Robustly Solving Sequential Tasks with Dynamically Composed Gradient Descent

Vito Mengers, Oliver Brock

Optimization

🎯 What it does: Propose a method for solving sequential tasks based on gradient descent with dynamic adjustment of potential fields, capable of completing long-sequence tasks without planning.

Non-Conservative Obstacle Avoidance for Multi-Body Systems Leveraging Convex Hulls and Predicted Closest Points

Lotte Rassaerts, Elena Torta

OptimizationRobotic Intelligence

🎯 What it does: Propose integrating future closest point prediction into collision avoidance controllers, utilizing convex hulls and closest point distance calculations.

Non-Destructive 3D Root Structure Modeling

Guoyu Lu

Convolutional Neural NetworkPoint CloudAgriculture Related

🎯 What it does: Proposed a method based on deep convolutional neural networks for target signal detection and curve parameter regression in multi-B-scan ground-penetrating radar (GPR) data, generating fitting curves to represent underground root structures, and designing a shape reconstruction network to reconstruct complete 3D root models from sparse slice 3D points.

Non-Parametric GNSS Integer Ambiguity Estimation via Positional Likelihood Field Marginalization

Aoki Takanose, Masashi Yokozuka

OptimizationSimultaneous Localization and MappingPhysics Related

🎯 What it does: Propose a non-parametric method to estimate the posterior distribution of GNSS integer ambiguities by defining a position information likelihood field in the position space and performing marginalization.

Non-Prehensile Shape Manipulation of Elastoplastic Objects With Reinforcement Learning

Sverre Herland, E. Misimi

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a non-grasping shape manipulation framework based on deep reinforcement learning, which deforms deformable objects into target shapes using continuously parameterized push actions.

Nonlinear Motion-Guided and Spatio-Temporal Aware Network for Unsupervised Event-Based Optical Flow

Zuntao Liu, Zheng Fang

Optical FlowSequential

🎯 What it does: Proposed an unsupervised event camera optical flow network called E-NMSTFlow for long-term temporal sequences, leveraging spatiotemporal information and nonlinear motion compensation to improve optical flow estimation.

NYC-Event-VPR: A Large-Scale High-Resolution Event-Based Visual Place Recognition Dataset in Dense Urban Environments

Tai-Yu Pan, Chen Feng

RecognitionRetrievalImageBenchmark

🎯 What it does: Proposed the NYC-Event-VPR dataset and evaluated its generalization performance using three frameworks

Obstacle-Avoidant Leader Following with a Quadruped Robot

C. Scheidemann, Marco Hutter

Robotic IntelligenceImageMultimodalityPoint Cloud

🎯 What it does: Proposed a virtual belt mechanism enabling robots to naturally follow operators, and employed a customized local obstacle avoidance planner for navigation in dynamic narrow environments; achieved tracking and obstacle avoidance through sensor fusion, and verified robustness and performance on the ANYmal platform.

Occ-LLM: Enhancing Autonomous Driving with Occupancy-Based Large Language Models

Tianshuo Xu, Yingcong Chen

Autonomous DrivingLarge Language ModelAuto Encoder

🎯 What it does: Proposed an LLM based on occupancy representation to enhance autonomous driving performance

Occlusion-Aware 6D Pose Estimation with Depth-Guided Graph Encoding and Cross-Semantic Fusion for Robotic Grasping

Jingyang Liu, Chenguang Yang

Pose EstimationRobotic IntelligenceGraph Neural NetworkImageMultimodality

🎯 What it does: Proposed a 6D pose estimation framework for occlusion scenarios, which utilizes a depth-guided graph neural network (GNN) to model latent relationships in RGBD inputs, and adaptively fuses two types of semantic information (mask and binary code) to extract 2D-3D correspondence features.

OccRWKV: Rethinking Efficient 3D Semantic Occupancy Prediction with Linear Complexity

Junming Wang, Qian Zhang

Autonomous DrivingComputational EfficiencyRobotic IntelligenceTransformerPoint Cloud

🎯 What it does: Proposed an efficient 3D semantic occupancy prediction network called OccRWKV, specifically designed for real-time 3D scene reconstruction in robotic navigation and autonomous driving systems.

OCCUQ: Exploring Efficient Uncertainty Quantification for 3D Occupancy Prediction

Severin Heidrich, Lutz Eckstein

Autonomous Driving

🎯 What it does: An efficient uncertainty quantification method is proposed for the 3D occupancy prediction task, evaluated under different camera failure scenarios.

ODYSSEE: Oyster Detection Yielded by Sensor Systems on Edge Electronics

Xiao-sheng Lin, Y. Aloimonos

Object DetectionData SynthesisConvolutional Neural NetworkDiffusion modelImageAgriculture Related

🎯 What it does: Using Stable Diffusion to generate realistic synthetic images, enhancing the collected real dataset, training the YOLOv10 visual model for oyster detection, and finally deploying and testing on edge platforms such as Aqua2 AUV

Off-Road Freespace Detection with LiDAR-Camera Fusion and Self-Distillation

Shuo Gu, Ming Gao

SegmentationAutonomous DrivingKnowledge DistillationMultimodalityPoint Cloud

🎯 What it does: Proposed a lightweight end-to-end freespace detection network using cascaded LiDAR-camera fusion and multi-scale self-distillation

Offline Adaptation of Quadruped Locomotion Using Diffusion Models

Reece O'Mahoney, Ioannis Havoutis

Robotic IntelligenceDiffusion model

🎯 What it does: Propose a diffusion model-based quadruped gait control framework that can interpolate between multiple skills and achieve offline adaptation to new gait behaviors after training, while being capable of running on the robot's local CPU.

OG-Gaussian: Occupancy Based Street Gaussians for Autonomous Driving

Yedong Shen, Yanyong Zhang

Autonomous DrivingGaussian SplattingImagePoint Cloud

🎯 What it does: Propose the OG-Gaussian method, using placeholder grids instead of LiDAR point clouds to achieve 3D scene reconstruction

OLiVia-Nav: An Online Lifelong Vision Language Approach for Mobile Robot Social Navigation

Siddarth Narasimhan, G. Nejat

Knowledge DistillationRobotic IntelligenceVision Language ModelMultimodality

🎯 What it does: Proposed an online lifelong vision-language architecture OLiVia-Nav for social navigation in mobile robots, with experimental validation in real-world environments.

Olympus: A Jumping Quadruped for Planetary Exploration Utilizing Reinforcement Learning for In-Flight Attitude Control

Jørgen Anker Olsen, Kostas Alexis

Robotic IntelligenceReinforcement Learning

🎯 What it does: This paper designs and simulates a jumping quadruped robot named Olympus, optimized for the low-gravity environment of Mars, to improve vertical jumping height, forward jumping distance, and achieve flight attitude reorientation; subsequently, a reinforcement learning-based flight attitude control strategy is proposed and trained, successfully bridging the gap from simulation to the real world; the effectiveness of this strategy is verified through experiments.

OMASTAR Optimal Magnetic Actuation System Arrangement

Veerash Palanichamy, Onaizah Onaizah

OptimizationPhysics Related

🎯 What it does: Designed a permanent magnet-driven magnetic system with a volume of 40 cm³, suitable for placement beneath an operating table, and proposed an E-optimal magnetic arrangement design method based on set function maximization.

OmniShape: Zero-Shot Multi-Hypothesis Shape and Pose Estimation in the Real World

Katherine Liu, Rares Ambrus

Pose EstimationDiffusion modelImage

🎯 What it does: Proposed a method called OmniShape for model-free, class-agnostic estimation of object pose and complete shape from a single observed image.

On Chain Driven, Adaptive, Underactuated Fingers for the Development of Affordable, Robust Humanlike Prosthetic Hands

Trevor Heinemann, Minas Liarokapis

Robotic Intelligence

🎯 What it does: Designed and verified a chain-driven, adaptive, underactuated finger for manufacturing affordable and durable humanoid prosthetic hands.

On the Benefits of Hysteresis in Tendon Driven Continuum Robots

David Hanley, Mohsen Khadem

Robotic Intelligence

🎯 What it does: Proposed and validated a hysteresis model applicable to tension-driven continuum robots.

On the Necessity of Real-Time Principles in GPU-Driven Autonomous Robots

Syed W. Ali, Ron Alterovitz

Computational EfficiencyRobotic Intelligence

🎯 What it does: Investigated the necessity of real-time principles in GPU-driven autonomous robots, applied the TimeWall framework to the computational components of autonomous drones, and experimentally validated the ability to maintain timeliness and safe operation even in the presence of interfering processes.

On the Synthesis of Reactive Collision-Free Whole-Body Robot Motions: A Complementarity-Based Approach

Haowen Yao, Sami Haddadin

OptimizationRobotic Intelligence

🎯 What it does: Proposed a Fast Linear Quadratic Compensation (FLIQC) motion planner for generating motion plans in high-degree-of-freedom systems that consider whole-body static and dynamic collisions, capable of real-time response to moving obstacles.

On-Robot Reinforcement Learning with Goal-Contrastive Rewards

Ondrej Biza, L. Wong

Robotic IntelligenceReinforcement LearningContrastive LearningVideo

🎯 What it does: Proposes Goal-Contrastive Rewards (GCR) method, which learns dense reward functions from passive video demonstrations and combines implicit value loss with goal contrastive loss to enable reinforcement learning for robots in real environments.

One Map to Find Them All: Real-time Open-Vocabulary Mapping for Zero-shot Multi-Object Navigation

F. Busch, Olov Andersson

Robotic IntelligenceVision-Language-Action ModelSimultaneous Localization and MappingBenchmark

🎯 What it does: Propose reusable open-vocabulary feature maps and probabilistic semantic map update methods for real-time object search, and create a zero-shot multi-target navigation benchmark that enables robots to leverage prior search information to improve search efficiency.

One Net to Rule Them All: Domain Randomization in Quadcopter Racing Across Different Platforms

Robin Ferede, G. D. de Croon

Domain AdaptationAutonomous DrivingRobotic Intelligence

🎯 What it does: Trained and validated a neural network controller that directly computes motor commands based solely on the current state, achieving high-performance control on racing drones of different sizes (3 inches and 5 inches) through domain randomization.

One-Shot Dual-Arm Imitation Learning

Yilong Wang, Edward Johns

Robotic IntelligenceMeta Learning

🎯 What it does: Developed One-Shot Dual-Arm Imitation Learning (ODIL), enabling dual-arm robots to learn precise and coordinated daily tasks with just a single demonstration.

One-Shot Imitation Under Mismatched Execution

K. Kedia, Sanjiban Choudhury

Data SynthesisRobotic IntelligenceVideoSequentialRetrieval-Augmented Generation

🎯 What it does: Proposes the RHyME framework, which automatically matches human and robot trajectories using a sequence-level optimal transport function, and synthesizes semantically equivalent human videos by retrieving and combining short-term human segments, enabling strategy training without paired data.

One-Shot Manipulation Strategy Learning by Making Contact Analogies

Yuyao Liu, L. Kaelbling

Robotic Intelligence

🎯 What it does: Propose the MAGIC method for one-time learning of operational strategies, utilizing reference action trajectories to match contact points and replicate strategies on new objects

One-Shot Video Imitation via Parameterized Symbolic Abstraction Graphs

Jianren Wang, Christopher G. Atkeson

Robotic IntelligenceGraph Neural NetworkWorld ModelVideo

🎯 What it does: Propose a method for learning the manipulation of dynamic and deformable objects from a single demonstration video using a parameterized symbolic abstract graph (PSAG), and apply it to real robot experiments.

Online Aggregation of Trajectory Predictors

Alex Tong, Heng Yang

Autonomous DrivingOptimizationMixture of ExpertsTime Series

🎯 What it does: Proposes a lightweight, model-agnostic online aggregation method that treats each trajectory predictor as an expert and maintains a probability vector to mix their outputs.

Online Design Optimization of Passive Exoskeletons Using Fast Biomechanics Simulation and Reinforcement Learning

Vighnesh Vatsal

OptimizationRobotic IntelligenceReinforcement LearningWorld ModelBiomedical DataPhysics Related

🎯 What it does: In dynamic tasks (arm extension and walking), MuJoCo biomechanical simulation and reinforcement learning (RL) are used to online optimize passive exoskeleton design parameters, aiming to reduce muscle load.

Online Diffusion-Based 3D Occupancy Prediction at the Frontier with Probabilistic Map Reconciliation

Alec Reed, Christoffer Heckman

Autonomous DrivingComputational EfficiencyDiffusion model

🎯 What it does: Implementing real-time online occupancy prediction using an improved diffusion model

Online Identification of Skidding Modes with Interactive Multiple Model Estimation

Ameya Salvi, V. Krovi

Robotic Intelligence

🎯 What it does: Proposes a filtering framework based on the Interactive Multiple Model (IMM) for online probabilistic identification of predefined robot operational modes, particularly slip modes caused by different terrains or loss of wheel traction.

Online Informative Motion Planning for Active Information Gathering of a Non-Stationary Gaussian Process

Kexiang Mao, Xiaoming Duan

OptimizationPhysics Related

🎯 What it does: Proposed an online information acquisition motion planning method for non-stationary Gaussian processes, including an information path planner, an adaptive speed planner, and a path smoothing and tracking strategy.

Online Multi-Robot Federated Learning for Distributed Coverage Control of Unknown Spatial Processes

Mattia Mantovani, Lorenzo Sabattini

Federated LearningRobotic Intelligence

🎯 What it does: Propose a federated learning framework for multi-robot distributed coverage control, which does not share raw data when training Gaussian process models and introduces a filtering strategy to select relevant samples

Online Nonlinear MPC for Multimodal Locomotion

Saverio Taliani, Daniele Pucci

OptimizationRobotic Intelligence

🎯 What it does: Proposes an online nonlinear MPC controller and its predictive model to stabilize the trajectory of a humanoid aerial robot in multimodal scenarios (walking and flying), and handle various walking strategies and transition actions.

Online Risk-Bounded Graph-Based Local Planning for Autonomous Driving With Theoretical Guarantees*

Abdulrahman Ahmad, A. Sumaiti

Autonomous DrivingOptimization

🎯 What it does: Proposed an online graph-based local planning method that uses a user-defined risk budget Δ to constrain the entire task, ensuring the vehicle navigates continuously, safely, and in real-time in dynamic environments;

Online Trajectory Replanner for Dynamically Grasping Irregular Objects

M. Vu, Anh Nguyen

OptimizationRobotic Intelligence

🎯 What it does: Proposed and implemented an online trajectory replanning framework for dynamic grasping of irregular objects.

Online Waypoint Recognition of Controlled Agents in Uncertain Environments

Jia Guo, Sarah Keren

Autonomous DrivingSimultaneous Localization and Mapping

🎯 What it does: Proposed an online waypoint recognition algorithm (OWR) that utilizes a Kalman filter to real-time identify the target waypoints of a controlled agent based on its observed behaviors and known dynamic models.

OoDIS: Anomaly Instance Segmentation and Detection Benchmark

Alexey Nekrasov, Matthias Rottmann

Object DetectionSegmentationAnomaly DetectionBenchmark

🎯 What it does: Expand existing OOD anomaly segmentation benchmarks by adding instance segmentation and object detection tasks, and provide corresponding evaluation and competition platforms.

Open-Loop Position Control of a Miniature Magnetic Robot Using Two-Dimensional Divergence Control of a Magnetic Force

Hakjoon Lee, Hongsoo Choi

Robotic IntelligencePhysics Related

🎯 What it does: Proposed an open-loop position control method for magnetic microrobots using two-dimensional magnetic force divergence control with fixed electromagnets.

Open-Nav: Exploring Zero-Shot Vision-and-Language Navigation in Continuous Environment with Open-Source LLMs

Yanyuan Qiao, Qi Wu

Autonomous DrivingTransformerLarge Language ModelVision Language ModelMultimodalityChain-of-Thought

🎯 What it does: Explore the use of open-source large language models for zero-shot vision-language navigation in continuous environments

Open-RGBT: Open-Vocabulary RGB-T Zero-Shot Semantic Segmentation in Open-World Environments

Meng Yu, Yufeng Yue

SegmentationPrompt EngineeringVision Language ModelMultimodality

🎯 What it does: Proposed the Open-RGBT model, achieving open-vocabulary semantic segmentation in RGB-T images through visual prompts and CLIP image-text similarity for instance-level detection.

Open-Set Semantic Uncertainty Aware Metric-Semantic Graph Matching

Kurran Singh, John J. Leonard

Object DetectionObject TrackingAutonomous DrivingTransformerImagePoint Cloud

🎯 What it does: Compute semantic uncertainty metrics and integrate them into an object-level uncertainty tracking framework to achieve robust object-level closed-loop detection for unknown object categories in marine scenarios; model the closed-loop detection problem as graph matching and validate in real-time underwater environments using a graph editing problem solver.

Open3DTrack: Towards Open-Vocabulary 3D Multi-Object Tracking

Ayesha Ishaq, R. Anwer

Object Tracking

🎯 What it does: Propose an open vocabulary 3D multi-object tracking framework

OpenBench: A New Benchmark and Baseline for Semantic Navigation in Smart Logistics

Junhui Wang, Guyue Zhou

Autonomous DrivingTransformerLarge Language ModelVision Language ModelTextMultimodalityBenchmark

🎯 What it does: Proposed the OPEN system and OpenBench benchmark, using OpenStreetMap as the map representation, combining large language models (LLMs) and vision-language models (VLMs) to achieve outdoor semantic navigation; experiments were conducted in simulation and real-world environments.

OpenGS-SLAM: Open-Set Dense Semantic SLAM with 3D Gaussian Splatting for Object-Level Scene Understanding

Dianyi Yang, Mengyin Fu

Gaussian SplattingSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposes the OpenGS-SLAM framework, which utilizes 3D Gaussian expansion for dense semantic SLAM in open set environments, integrating explicit semantic labels from 2D base models into the 3D Gaussian framework, achieving fast 2D label map rendering, scene updates, and robust 3D object-level scene understanding.

OpenSU3D: Open World 3D Scene Understanding Using Foundation Models

Rafay Mohiuddin, André Borrmann

RecognitionRepresentation LearningTransformerLarge Language ModelVision Language ModelPoint CloudMesh

🎯 What it does: Propose an scalable method that leverages 2D foundation models to incrementally construct open-set, instance-level 3D scene representations, and efficiently aggregates instance-level masks, feature vectors, names, and descriptions; introduce a feature vector fusion scheme to enhance contextual knowledge and explore large language models for automatic annotation and spatial reasoning.

OPPA: Online Planner's Parameter Adaptation for Enhanced Mobile Robot Navigation

Minsu Chang, Hyundo Choi

OptimizationRobotic IntelligenceTransformerPoint Cloud

🎯 What it does: Proposed the OPPA framework, which enhances the adaptability and safety of mobile robots in dynamic or unstructured environments through dynamic adjustment of planner parameters.

OPRNet: Object-Centric Point Reconstruction Network for Multimodal 3D Object Detection in Adverse Weathers

Jae Hyun Yoon, S. Yoo

Object DetectionMultimodalityPoint Cloud

🎯 What it does: Propose a point reconstruction network based on isometric projection for multi-modal 3D object detection, which includes a range-constrained noise filter, an object-oriented point generator, and a dual 2D auxiliary module.

Optimal Fault-Tolerant Control for Tugboats Robust Path Following in Nearshore

Jiangteng Shi, Jia Ren

OptimizationRobotic Intelligence

🎯 What it does: Proposed an optimal fault-tolerant control scheme for autonomous tugs to achieve robust path tracking.

Optimal Framework for Constrained Admittance Path-Following Control

Giulio Besi, F. Ferraguti

Optimization

🎯 What it does: Proposed an optimal controller for constrained admittance control, achieving strict adherence to constraint boundaries while minimizing kinematic energy variation

Optimal Gait Control for a Tendon-Driven Soft Quadruped Robot by Model-Based Reinforcement Learning

Xuezhi Niu, Lei Feng

Robotic IntelligenceReinforcement Learning

🎯 What it does: A model-based reinforcement learning (MBRL) method is proposed for optimal gait control of a soft quadruped robot equipped with four compressible traction-driven soft actuators.

Optimal Motion Planning for a Class of Dynamical Systems

Panagiotis Rousseas, Kostas J. Kyriakopoulos

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a new method for optimal motion planning for a certain class of dynamic systems

Optimal Torque Distribution via Dynamic Adaptation for Quadrupedal Locomotion on Slippery Terrains

Despina Ekaterini Argiropoulos, P. Trahanias

OptimizationRobotic Intelligence

🎯 What it does: Propose a real-time adaptive quadrupedal robot locomotion controller that maintains stability and controllability on highly slippery surfaces.

Optimal Trajectory Planning for Cooperative Manipulation with Multiple Quadrotors Using Control Barrier Functions

Arpan Pallar, Giuseppe Loianno

OptimizationRobotic Intelligence

🎯 What it does: Proposed a trajectory planning algorithm for multirotor collaborative transportation using Control Barrier Functions (CBF).

Optimization-Based Task and Motion Planning Under Signal Temporal Logic Specifications Using Logic Network Flow

Xuan Lin, Ye Zhao

Optimization

🎯 What it does: Proposes an optimization-based task and motion planning framework called 'Logic Network Flow,' which integrates Signal Temporal Logic (STL) specifications into Mixed Integer Linear Programming.

Optimize and Coordinate Multiple DMPs Under Constraints to Achieve a Collaborative Manipulation Task

Ali H. Kordia, Francisco S. Melo

OptimizationRobotic Intelligence

🎯 What it does: Studies how to collaboratively execute and optimize pre-learned dynamic movement primitives (DMP) in multi-robot/multi-robotic arm environments to accomplish collaborative tasks.

Optimizing Complex Control Systems with Differentiable Simulators: A Hybrid Approach to Reinforcement Learning and Trajectory Planning

Amit Parag, E. Misimi

OptimizationRobotic IntelligenceReinforcement LearningWorld Model

🎯 What it does: Combine learning with a differentiable simulator to enhance exploration-exploitation efficiency in deep reinforcement learning, learning value functions, state trajectories, and control strategies, and utilizing locally optimal model trajectory optimization results.

Optimizing Efficiency of Mixed Traffic Through Reinforcement Learning: A Topology-Independent Approach and Benchmark

Chuyang Xiao, Yuexin Ma

Autonomous DrivingOptimizationReinforcement LearningBenchmark

🎯 What it does: Proposed a model-agnostic reinforcement learning control strategy for mixed traffic flow, and released a real-world benchmark containing 444 scenarios.

Optimizing NeRF-Based SLAM with Trajectory Smoothness Constraints

Yicheng He, Hong Zhang

Neural Radiance FieldSimultaneous Localization and Mapping

🎯 What it does: Propose TS-SLAM, which incorporates smoothness constraints based on a unified cubic B-spline into the camera trajectory to achieve continuous acceleration and ensure smooth motion; leverage the differentiability and local control properties of B-splines for end-to-end incremental learning of control points within a sliding window; simultaneously utilize dynamic priors to further smooth the trajectory.

Optimizing Robot Programming: Mixed Reality Gripper Control

Maximilian Rettinger, Gerhard Rigoll

OptimizationRobotic Intelligence

🎯 What it does: Conducted comparative experiments using three controller-based robot programming methods in a mixed reality environment.

Optimizing Underwater Robot Navigation: A Study of DRL Algorithms and Multi-Modal Sensor Fusion

Md. Ether Deowan, Ricard Marxer

Depth EstimationRobotic IntelligenceReinforcement LearningMultimodality

🎯 What it does: This paper studies the integration of multiple reinforcement learning algorithms (PPO, TRPO, SAC, TD3, A2C) with multi-modal sensor fusion to enhance the autonomous navigation capabilities of low-cost underwater robots in complex environments.

ORB-SfMLearner: ORB-Guided Self-supervised Visual Odometry with Selective Online Adaptation

Yanlin Jin, Yuzhong Zhong

Pose EstimationDomain AdaptationAutonomous DrivingConvolutional Neural NetworkTransformerSimultaneous Localization and MappingImageVideo

🎯 What it does: Propose a self-supervised visual odometry method called ORB-SfMLearner based on ORB features, and introduce selective online adaptation technology

ORLA*: Mobile Manipulator-Based Object Rearrangement with Lazy A*

Kai Gao, Jingjin Yu

OptimizationRobotic Intelligence

🎯 What it does: Proposes ORLA*, a mobile manipulator object reordering planning method that utilizes delayed/lazy evaluation search, capable of considering end-effector and robot base movement, handling multi-layer reordering tasks, and achieving time-optimal pick-and-place sequences.

Overlapping Free: Anchorless UWB-Assisted Relative Pose Estimation for Multi-Robot Systems

Yanpu Yun, Danwei Wang

Pose EstimationOptimizationRobotic IntelligenceSimultaneous Localization and MappingMultimodalityPoint Cloud

🎯 What it does: Proposes the Anchorless UWB-Assisted Relative Pose Estimation (AURPE) method for relative pose estimation in multi-robot systems under non-overlapping fields of view.

Overlapping Social Navigation Principles: A Framework for Social Robot Navigation

Bryce Ikeda, Christina Soyoung Songa

Robotic IntelligenceVideo

🎯 What it does: Implemented autonomous robots using SRN behaviors in scenarios, and validated the framework through online experiments where participants ranked videos of different SRN behavior combinations.

P3-PO: Prescriptive Point Priors for Visuo-Spatial Generalization of Robot Policies

Mara Levy, Abhinav Shirivastava

Representation LearningRobotic IntelligenceReinforcement LearningImage

🎯 What it does: Proposed the P3-PO framework, which constructs unique state representations by leveraging human-annotated semantic points and utilizing visual model propagation to enhance the generalization ability of robotic manipulation.

PACE: Proactive Assistance in Human-Robot Collaboration Through Action-Completion Estimation

D. D. Lazzari, D. Romeres

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement Learning

🎯 What it does: Proposes the PACE proactive assistance framework based on action completion estimation, aiming to improve human-robot collaboration efficiency through real-time monitoring of human progress.

Panoptic-Depth Forecasting

Juana Valeria Hurtado, Abhinav Valada

SegmentationDepth EstimationAutonomous DrivingTransformerImageVideo

🎯 What it does: Propose a panoptic-depth prediction task for future frames and construct corresponding datasets and evaluation metrics.

Parallel-Constraint Model Predictive Control: Exploiting Parallel Computation for Improving Safety

Elias Fontanari, Andrea Del Prete

OptimizationSafty and Privacy

🎯 What it does: Implemented Parallel-Constraint Model Predictive Control (MPC), solving multiple MPC problems simultaneously, where each problem instantiates safety set constraints at different time steps within the prediction horizon and selects the optimal solution based on predefined criteria;

Parking-SG: Open-Vocabulary Hierarchical 3D Scene Graph Representation for Open Parking Environments

Yaowen Zhang, Mengyin Fu

Autonomous DrivingRepresentation LearningGraph Neural NetworkWorld ModelGraph

🎯 What it does: Propose Parking-SG, an open-vocabulary hierarchical 3D scene graph representation for open and complex parking environments, supporting autonomous valet parking.

Partial-to-Full Registration based on Gradient-SDF for Computer-Assisted Orthopedic Surgery

Tiancheng Li, Shoudong Huang

🎯 What it does: Proposes a gradient SDF-based partial-to-complete registration framework for bone registration in computer-assisted surgery.

Passivity Filters for Bilateral Teleoperation with Variable Impedance Control

Fadi Alyousef Almasalmah, Bernard Bayle

OptimizationSafty and PrivacyRobotic Intelligence

🎯 What it does: Proposed an optimization-based bidirectional teleoperation variable damping control passivity-filtering framework, integrating three techniques: PF, TDP, and PSPM;

Patch Tree: Exploiting the Gauss Map and Principal Component Analysis for Robotic Grasping

Yan-Bin Jia, Ling Tang

Robotic Intelligence

🎯 What it does: Segment the object surface, construct a Patch Tree, and use it for grasping planning and optimization, with simulation and experiments conducted on Shadow Hand

Path Planning Using Instruction-Guided Probabilistic Roadmaps

Jiaqi Bao, Ryo Yonetani

Autonomous DrivingOptimizationTransformerLarge Language ModelText

🎯 What it does: Propose a data-driven path planning algorithm named IG-PRM

Pedestrian Intention and Trajectory Prediction in Unstructured Traffic Using IDD-PeD

Ruthvik Bokkasam, C. V. Jawahar

Autonomous DrivingVideoBenchmark

🎯 What it does: Studied pedestrian intent and trajectory prediction in autonomous driving environments, proposed the IDD-PeD dataset, and evaluated existing prediction methods.

Perfectly Undetectable False Data Injection Attacks on Encrypted Bilateral Teleoperation System based on Dynamic Symmetry and Malleability

Hyukbin Kwon, Jun Ueda

Adversarial AttackRobotic IntelligencePhysics Related

🎯 What it does: Investigate the vulnerability of bidirectional teleoperation systems to completely undetectable false data injection attacks (FDIAs), and demonstrate the attack methods through experiments

Persistent Object Gaussian Splat (POGS) for Tracking Human and Robot Manipulation of Irregularly Shaped Objects

Justin Yu, Ken Goldberg

Pose EstimationDepth EstimationRobotic IntelligenceGaussian SplattingImagePoint Cloud

🎯 What it does: Propose the Persistent Object Gaussian Splat (POGS) system, which achieves persistent pose estimation for irregularly shaped and previously unseen objects using semantic, self-supervised visual features, and object grouping features, enabling grasp, reorientation, and natural language-driven manipulation without requiring expensive rescan or CAD models.

Personalization in Human-Robot Interaction Through Preference-Based Action Representation Learning

Ruiqi Wang, Byung-Cheol Min

Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningTabularBenchmark

🎯 What it does: proposes PbARL, an efficient fine-tuning method that learns action representations by leveraging pre-trained robot policies and maximizing mutual information between the source domain and the target user preference-aligned domain, thereby decoupling task structure from preferences;

Personalizing Interfaces to Humans with User-Friendly Priors

Benjamin A. Christie, Dylan P. Losey

Robotic Intelligence

🎯 What it does: Enable robots to personalize their interface signals according to the current user, aligning human interpretation with the robot's intent.

Physical Simulation with Force Feedback Aids Robot Factors Design

Carina Kaeser, Justin Werfel

Robotic IntelligenceWorld ModelPhysics Related

🎯 What it does: Developed a 3D physics simulation tool based on Unity, combined with Phantom Omni / Geomagic Touch haptic devices, to evaluate whether the device design is easy for robotic operation.

Physically-Consistent Parameter Identification of Robots in Contact

Shahram Khorshidi, Majid Khadiv

Robotic IntelligencePhysics Related

🎯 What it does: A method for robot inertia parameter identification is proposed using only joint current/torque measurements, without the need for contact force measurements.

Physics-Aware Robotic Palletization With Online Masking Inference

Tianqi Zhang, Wei Zhan

Robotic IntelligenceReinforcement LearningPhysics Related

🎯 What it does: Proposed a box stacking planning method based on reinforcement learning, using online action space mask inference to guide the policy in selecting effective actions.

Physics-Informed Hybrid Modeling of Pneumatic Artificial Muscles

Ge Wang, M.T. Pham

Physics Related

🎯 What it does: The study integrates physical models with neural networks to perform hybrid modeling of PAM.

Physics-Informed Split Koopman Operators for Data-Efficient Soft Robotic Simulation

Eron Ristich (Arizona State University), Jiefeng Sun (Arizona State University)

Robotic IntelligencePhysics Related

🎯 What it does: Propose a physics-informed Koopman operator identification method for achieving soft robot simulation under small datasets, validated on a traction-driven soft robotic arm.

PhysPart: Physically Plausible Part Completion for Interactable Objects

Rundong Luo, Siyuan Huang

GenerationDiffusion modelPhysics Related

🎯 What it does: Propose a physics-feasible part completion method based on diffusion models, utilizing geometric conditioning and classifier-free guidance, and guiding sampling through stability and mobility loss to generate parts that precisely fit objects and move smoothly.

Pinto: A Latched Spring Actuated Robot for Jumping and Perching

Christopher Y. Xu, Justin K. Yim

Robotic Intelligence

🎯 What it does: Designed and implemented a 450g 'Pinto' miniature robot capable of jumping from the ground to vertical tree trunks like a squirrel, equipped with a 5-bar leg mechanism that can switch between different elastic modes and a dual-degree-of-freedom telescoping arm with a spiny gripper for rapid branch landing.

PIP-Loco: A Proprioceptive Infinite Horizon Planning Framework for Quadrupedal Robot Locomotion

Aditya Shirwatkar, Shishir N Y Kolathaya

Robotic IntelligenceReinforcement Learning

🎯 What it does: Designed and verified a quadruped robot locomotion planning framework PIP-Loco that combines reinforcement learning with an internal model (MPC), enabling flexible and safe long-term planning and control.