arXivSub Start free trial

IROS 2025 Papers — Page 4

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1984 papers

BoRe-Depth: Self-Supervised Monocular Depth Estimation with Boundary Refinement for Embedded Systems

Chang Liu, Xu Zhang

Depth EstimationImage

🎯 What it does: Proposed the BoRe-Depth lightweight monocular depth estimation model, achieving efficient depth image estimation and boundary refinement on embedded systems.

Botany-Bot: Digital Twin Monitoring of Occluded and Underleaf Plant Structures with Gaussian Splats

S. Adebola, Kenneth Y. Goldberg

Robotic IntelligenceGaussian SplattingAgriculture Related

🎯 What it does: Construct a detailed 'annotated digital twin' plant model and develop robotic algorithms to manipulate leaves for obtaining high-resolution indexable images of occluded regions.

Bounding Distributional Shifts in World Modeling through Novelty Detection

Eric Jing, Abdeslam Boularias

Robotic IntelligenceAuto EncoderWorld Model

🎯 What it does: Propose using a Variational Autoencoder (VAE) as a novelty detector to ensure that action trajectories generated in model predictive control do not cause the learned world model to deviate from the training data distribution, thereby enhancing the robustness of planning-based models.

Braking Control in Clutched-Elastic Robots: Coordinating the Underactuation-to-Actuation Transition

Vasilije Rakcevic, Sami Haddadin

Robotic Intelligence

🎯 What it does: We propose a feedback-based two-phase method for safe braking in the clutch state of an elastic robotic arm.

Bridge the Gap: Enhancing Quadruped Locomotion with Vertical Ground Perturbations

Maximilian Stasica, André Seyfarth

Domain AdaptationRobotic IntelligenceReinforcement Learning

🎯 What it does: Training the Unitree Go2 quadruped robot using the PPO algorithm in the MuJoCo simulation environment to achieve a stable gait on an oscillating bridge with a frequency of 2.0 Hz, and realize zero-shot transfer to a real bridge.

Bridging Text and Vision: A Multi-View Text-Vision Registration Approach for Cross-Modal Place Recognition

Tianyi Shang, Weijun Hu

RecognitionRetrievalTransformerLarge Language ModelVision Language ModelImageTextMultimodality

🎯 What it does: Proposed a multi-view text-visual registration method called Text4VPR for cross-modal place recognition, achieving the first-ever matching between text descriptions and image databases using only text.

Bridging the Reality Gap: Communication-Aware Task Allocation with Multi-Objective Asynchronous Policy Learning

Zehao Xiong, Jie Li

OptimizationReinforcement Learning

🎯 What it does: Proposes a communication-aware task allocation method for UAV swarms, trains a gating mechanism strategy to coordinate communication timing, and achieves collaborative minimization of task conflicts and communication costs through POMDP modeling, asynchronous experience collection and concatenation, and MOCPPO multi-objective optimization, ultimately verifying its superiority in HIL environments.

Budget-optimal multi-robot layout design for box sorting

Peiyu Zeng, Stelian Coros

Optimization

🎯 What it does: Proposes an optimization-based computational framework to automatically generate multi-robot layouts, minimize hardware budget, and abstract motion constraints through precomputed reachability graphs.

Building Hybrid Omnidirectional Visual-Lidar Map for Visual-Only Localization

Jingyang Huang, Ming Yang

Pose EstimationComputational EfficiencySimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposes a hybrid mapping and pure visual relocalization framework that integrates visual and LiDAR data, specifically designed for drones with limited computational resources.

Building Knowledge from Interactions: An LLM-Based Architecture for Adaptive Tutoring and Social Reasoning

Luca Garello, A. Sciutti

Robotic IntelligenceTransformerLarge Language ModelRetrieval-Augmented Generation

🎯 What it does: Developed an LLM-based robot coach agent that integrates a multimodal cognitive framework and an experience memory system to achieve a balance between social dialogue, task guidance, and goal-driven actions.

Bunny-VisionPro: Real-Time Bimanual Dexterous Teleoperation for Imitation Learning

Runyu Ding, Xiaolong Wang

Robotic Intelligence

🎯 What it does: This paper proposes Bunny-VisionPro, a system that realizes real-time dexterous teleoperation with both hands using a VR headset.

C-TRAC: Terrain-Adaptive Control for Articulated Tracked Robots via Contact-Aware Reinforcement Learning

Hainan Pan, Huimin Lu

Robotic IntelligenceReinforcement LearningAuto Encoder

🎯 What it does: Proposed and implemented a contact-aware reinforcement learning terrain-adaptive control framework (C-TRAC) for stable motion of articulated tracked robots on uneven terrain.

CA-W3D: Leveraging Context-Aware Knowledge for Weakly Supervised Monocular 3D Detection

Chupeng Liu, Weidong Cai

Object DetectionAutonomous DrivingKnowledge DistillationContrastive LearningImage

🎯 What it does: Propose the CA-W3D method, achieving weakly supervised monocular 3D detection through two-stage training: the pre-training stage employs Region-wise Object Contrastive Matching (ROCM) to align the 3D encoder with the region embeddings of the 2D visual localization model; the second stage uses Dual-to-One Distillation (D2OD) for pseudo-label training.

CA2Point: Learning Keypoint Detection and Description with Context Aggregation and Cross Augmentation

Xuebin Meng, Yinhe Han

Pose EstimationConvolutional Neural NetworkTransformer

🎯 What it does: Enhance keypoint detection and description through the interaction between global Transformer information and keypoints with descriptors.

CageCoOpt: Enhancing Manipulation Robustness through Caging-Guided Morphology and Policy Co-Optimization

Yifei Dong, Florian T. Pokorny

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a hierarchical framework called CageCoOpt that jointly optimizes the robotic gripper structure (morphology) and control strategy to achieve robust manipulation based on caging.

Cal or No Cal? - Real-Time Miscalibration Detection of LiDAR and Camera Sensors

Ilir Tahiraj, Markus Lienkamp

Anomaly DetectionAutonomous DrivingComputational EfficiencyContrastive LearningImagePoint Cloud

🎯 What it does: Proposed a real-time error detection framework based on contrastive learning, classifying calibration states into calibrated or misaligned through binary classification.

CalibMutiL: Online Calibration Of LiDAR-Camera Based On Multi-level Visual Feature Fusion

Guanghui Zhang, Askar Hamdulla

Autonomous DrivingConvolutional Neural NetworkImagePoint Cloud

🎯 What it does: Proposed a LiDAR-camera online calibration network called CalibMutiL based on multi-level visual feature fusion, capable of end-to-end alignment of point clouds and RGB images

Camera-tracked Soft Underwater Robot Enabling Robust Orientation Control for Maneuverability

Gabriele Bianchi, Josie Hughes

OptimizationRobotic IntelligenceVideo

🎯 What it does: Developed a control scheme for a soft robotic underwater vehicle based on a PD controller and real-time camera feedback, achieving continuous and reliable free-swimming control, capable of maintaining precise waypoint tracking for over 60 minutes, with a minimum turning radius of 27 centimeters, demonstrating high maneuverability relative to its volume.

CAMSCKF: A Multi-State Constraint Kalman Filter with Adaptive Multivariate Noise Parameters Clustering and Estimation for Visual-Inertial Odometry

Yiyang Tang, Yulong Huang

Autonomous DrivingOptimizationRobotic IntelligenceSimultaneous Localization and MappingVideo

🎯 What it does: Proposed a multi-state constrained Kalman filter (CAMSCKF) that utilizes adaptive multivariate noise parameter clustering and estimation to achieve real-time tracking and adjustment of the noise covariance matrix in visual-inertial odometry measurements.

Can Real-Time Lipreading Improve Speech Recognition? A Systematic Exploration Using Human-Robot Interaction Data

Sander Goetzee, Koen V. Hindriks

RecognitionRobotic IntelligenceVideoMultimodalityAudio

🎯 What it does: On the social robot Pepper platform, the system evaluated and compared the performance of offline and real-time audio-visual speech recognition (AVSR) models with audio-only models in laboratory and noisy public environments;

Can Real-to-Sim Approaches Capture Dynamic Fabric Behavior for Robotic Fabric Manipulation?

Yingdong Ru, Gerardo Aragon-Camarasa

Domain AdaptationRobotic IntelligencePhysics Related

🎯 What it does: Systematically evaluated three state-of-the-art Real-to-Sim parameter estimation methods (two micro-pipeline approaches and one data-driven method), proposed a novel physics-informed neural network (PINN) method for physical parameter estimation, combined these methods with two simulation engines, and evaluated them across multiple real-to-simulation scenarios (lifting, wind, stretching) with five different fabric types, while testing performance in three unseen scenarios (folding, throwing, swaying).

CAP: A Connectivity-Aware Hierarchical Coverage Path Planning Algorithm for Unknown Environments using Coverage Guidance Graph

Zongyuan Shen, Matthew Travers

OptimizationRobotic IntelligenceGraph

🎯 What it does: Proposed a connectivity-aware hierarchical coverage path planning algorithm CAP for efficient coverage in unknown environments.

CapsDT: Diffusion-Transformer for Capsule Robot Manipulation

Xiting He, Hongliang Ren

Robotic IntelligenceTransformerVision-Language-Action ModelDiffusion model

🎯 What it does: Propose the CapsDT model for gastric cyst robot operations, develop a magnetic-driven cyst robot system and corresponding simulator dataset, and evaluate multiple endoscopic tasks.

Capsizing-Guided Trajectory Optimization for Autonomous Navigation with Rough Terrain

Wei Zhang, Chaoqun Wang

Autonomous DrivingOptimization

🎯 What it does: Proposes an overturn-aware trajectory planning method for rough terrain, generating both safe and efficient trajectories.

CasPoinTr: Point Cloud Completion with Cascaded Networks and Knowledge Distillation

Yifan Yang, Jian Pu

RestorationKnowledge DistillationPoint Cloud

🎯 What it does: Proposes a point cloud completion framework called CasPoinTr, which uses a cascaded network and knowledge distillation to reconstruct missing parts of incomplete point clouds.

CATCH-FORM-3D: Compliance-Aware Tactile Control and Hybrid Deformation Regulation for 3D Viscoelastic Object Manipulation

Hongjun Ma, Weichang Li

Robotic IntelligenceMultimodalityOrdinary Differential Equation

🎯 What it does: Developed a precise contact force control and surface deformation regulation framework CATCH-FORM-3D for viscoelastic object manipulation.

Category-Level 6D Object Pose Estimation in Agricultural Settings Using a Lattice-Deformation Framework and Diffusion-Augmented Synthetic Data

Marios Glytsos, Petros Maragos

Data SynthesisPose EstimationDiffusion modelImageAgriculture Related

🎯 What it does: Proposed the PLANTPose framework, which can perform category-level 6D pose estimation using only RGB images, and adapts to the shapes of different instances by predicting deformation parameters;

Category-level Meta-learned NeRF Priors for Efficient Object Mapping

Saad Ejaz, J. L. Sánchez-López

Pose EstimationOptimizationMeta LearningNeural Radiance Field

🎯 What it does: Propose PRENOM, a prior-based efficient neural object mapper that combines category-level priors with object-level NeRF to achieve efficient reconstruction and canonical pose estimation.

Causal-Planner: Causal Interaction Disentangling with Episodic Memory Gating for Autonomous Planning

Yibo Yuan, Jianru Xue

Autonomous DrivingRecurrent Neural NetworkGraph Neural Network

🎯 What it does: Propose Causal-Planner, which uses attention adversarial graph learning to decouple causal and confounding factors in the scene interaction graph, and introduces the Long Short-Term Memory Gating Module (LSTEM) to enhance the capture of causal relationships in dynamic environments.

CCDP: Composition of Conditional Diffusion Policies with Guided Sampling

Amirreza Razmjoo, Fan Zhang

Robotic IntelligenceDiffusion model

🎯 What it does: Proposed a guided sampling strategy based on a conditional diffusion model, which avoids previous failed actions by gradually correcting the sampling distribution and infers recovery actions using only successful demonstration data;

CDIS : Cross-Dimensional Class-Agnostic 3D Instance Segmentation via 2D Mask Tracking and 3D-2D Projection Merging

Juno Kim, Byoung-Tak Zhang

Segmentation

🎯 What it does: Proposed a cross-dimensional, class-agnostic 3D instance segmentation framework called CDIS, which achieves globally consistent 3D instance labels through a feedback loop by tracking 2D instance masks and associating them with 3D superpoints, without requiring 3D-specific training.

CDP: Constrained Diffusion Policies with Mirror Diffusion Model for Safety-Assured Imitation Learning

Taeoh Ha, Daehyun Ji

Reinforcement Learning from Human FeedbackDiffusion model

🎯 What it does: Proposes a Constrained Diffusion Policy (CDP) framework to strictly adhere to safety constraints while imitating expert demonstrations.

Celebi’s Choice: Causality-Guided Skill Optimisation for Granular Manipulation via Differentiable Simulation

Minglun Wei, Ze Ji

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Propose a causality-guided skill optimization method named Celebi, combining differentiable physics simulation with causal inference-based adaptive step size adjustment for path optimization in soil particle-level operations.

CG-3DGS: Complexity-Guided 3D Gaussian Splatting for High-Fidelity Surgical Scene Reconstruction

Yao Yao, Cancan Zhao

RestorationGaussian SplattingOptical FlowVideoBiomedical Data

🎯 What it does: Proposed a complexity-guided 3D Gaussian Splatting (CG-3DGS) framework for high-fidelity surgical scene reconstruction.

CG-Net: Urban Trajectory Forecasting with Bipartite Graphs for Agents, Scene Context and Candidate Centerlines

Kaushik Bhowmik, Philippe Martinet

Autonomous DrivingGraph Neural NetworkTime SeriesSequential

🎯 What it does: Proposes the CandidateGraph-Net (CG-Net) framework for predicting the trajectories of target pedestrians or vehicles in urban intersection scenarios, by encoding candidate centerlines available at the current position and combining them with a bipartite graph attention network for interaction modeling.

CGS-SLAM: Compact 3D Gaussian Splatting for Dense Visual SLAM

Tianchen Deng, Weidong Chen

Gaussian SplattingSimultaneous Localization and Mapping

🎯 What it does: Propose a compact 3D Gaussian rendering SLAM system that reduces the number of Gaussian ellipsoids and parameter scale by utilizing a sliding window occlusion strategy and geometric codebook quantization technique, and achieves robust and accurate pose estimation through local-to-global bundle adjustment.

CHADET: Cross-Hierarchical-Attention for Depth-Completion Using Unsupervised Lightweight Transformer

Kevin Christiansen Marsim, Hyun Myung

Depth EstimationAutonomous DrivingConvolutional Neural NetworkTransformerImagePoint Cloud

🎯 What it does: Proposed a lightweight depth completion network named CHADET, which can generate accurate dense depth maps from RGB images and sparse depth points.

Chain-of-Imagination for Reliable Instruction Following in Decision Making

Enshen Zhou, Jing Shao

Robotic IntelligenceTransformerLarge Language ModelAgentic AIVision-Language-Action ModelDiffusion modelMultimodalityChain-of-Thought

🎯 What it does: Propose the Chain-of-Imagination mechanism and the DecisionDreamer low-level controller, enabling embodied agents to gradually imagine future states and execute based on context-aware visual subgoals, thereby improving reliable decision-making for text instructions.

ChatBuilder: LLM-assisted Modular Robot Creation

Xin Chen, Zherong Pan

Robotic IntelligenceTransformerLarge Language ModelAgentic AIText

🎯 What it does: Using LLM agents to automate the planning and assembly of modular robot structures.

CIT: Context-Based Biased Batch-Sampling for Almost-Surely Asymptotically Optimal Motion Planning

Liding Zhang, A. Knoll

Autonomous DrivingOptimization

🎯 What it does: Developed the CIT* sample-based motion planning algorithm, enhancing exploration efficiency through context-based biased sampling.

CLAIM: Camera-LiDAR Alignment with Intensity and Monodepth

Zhuo Zhang, Yikang Ding

Depth EstimationAutonomous DrivingImagePoint Cloud

🎯 What it does: This paper proposes the CLAIM method, which utilizes a powerful monocular depth model to achieve data alignment between the camera and LiDAR.

CLAP: A Closed-Loop Diffusion Transformer Action Foundation Model for Robotic Manipulation

Mu Li, Chenguang Yang

Robotic IntelligenceTransformerVision-Language-Action ModelDiffusion model

🎯 What it does: Proposed an advanced Vision-Language-Action (VLA) model by decomposing the model architecture into specialized action modules and critic modules, forming a closed-loop reasoning framework, and employing a diffusion action transformer to model continuous-time actions;

Class-Aware PillarMix: Can Mixed Sample Data Augmentation Enhance 3D Object Detection with Radar Point Clouds?

Miao Zhang, Bin Yang

Object DetectionAutonomous DrivingPoint Cloud

🎯 What it does: Proposed a hybrid sample data augmentation method for radar point clouds called Class-Aware PillarMix (CAPMix), which applies MixUp at the pillar level and uses class labels to guide the mixing ratio.

CLEA: Closed-Loop Embodied Agent for Enhancing Task Execution in Dynamic Environments

Mingcong Lei, Jinke Ren

Robotic IntelligenceTransformerLarge Language ModelAgentic AIMultimodality

🎯 What it does: Developed the Closed-Loop Embodied Agent (CLEA) architecture, utilizing four open-source LLMs for functional decoupling to achieve closed-loop task management;

Clevis and Tenon Assembly Using Visual Guiding Fields

Gilles Chabert, Adolfo Suarez-Roos

OptimizationRobotic IntelligenceOptical FlowImage

🎯 What it does: Proposed a visual servoing strategy for clevis and tenon joint assembly, using a vector field derived from images to guide motion, enabling the moving parts to move along an accurate 3D straight trajectory.

CLGA: A Collaborative LLM Framework for Dynamic Goal Assignment in Multi-Robot Systems

Xin Yu, Wenjun Wu

OptimizationRobotic IntelligenceTransformerLarge Language ModelTextBenchmark

🎯 What it does: Propose the CLGA framework, leveraging large language models (LLMs) for pre-planning tasks, with an external solver generating initial goal allocations, while small-scale models adjust in real-time during execution to address dynamic goal allocation problems in multi-robot systems; also released an NLP-based goal allocation benchmark dataset;

Closing the intent-to-behavior gap via Fulfillment Priority Logic

B. Mabsout, Renato Mancuso

OptimizationReinforcement Learning

🎯 What it does: Proposes the concept of goal fulfillment and constructs Fulfillment Priority Logic (FPL) to express intentions and priorities through logical formulas in multi-objective reinforcement learning; meanwhile designs the Balanced Policy Gradient algorithm to achieve better sample efficiency.

CM-LIUW-Odometry: Robust and High-Precision LiDAR-Inertial-UWB-Wheel Odometry for Extreme Degradation Coal Mine Tunnels

Kun Hu, Gongbo Zhou

Autonomous DrivingSimultaneous Localization and MappingMultimodalityPoint Cloud

🎯 What it does: Proposed a multi-modal SLAM framework CM-LIUW-Odometry based on Iterated Error-State Kalman Filter, integrating LiDAR-IMU-UWB and wheel encoders to achieve localization and mapping in complex underground coal mine environments.

Co-Adaptation of Embodiment and Control with Self-Imitation Learning

Sergio Hernández-Gutiérrez, K. Luck

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose a collaborative adaptation method that combines reinforcement learning with state-aligned self-imitation learning, enabling simultaneous optimization of a robot's body structure and behavioral policies within a limited number of design iterations.

Coarse-to-Fine Learning for Multi-Pipette Localisation in Robot-Assisted In Vivo Patch-Clamp

Lan Wei, Dandan Zhang

Object DetectionOptimizationRobotic IntelligenceTransformerGenerative Adversarial NetworkBiomedical Data

🎯 What it does: Propose a heatmap-enhanced coarse-to-fine learning method to achieve real-time localization for multi-tube tips.

COARSE: Collaborative Pseudo-Labeling with Coarse Real Labels for Off-Road Semantic Segmentation

Aurelio Noca, Deegan Atha

SegmentationDomain AdaptationAutonomous DrivingTransformerImage

🎯 What it does: Proposes the COARSE framework for semi-supervised domain adaptation in grounded semantic segmentation, combining coarse intra-domain labels with dense inter-domain labeled data.

Cockroach’s Turning Strategy Enhanced Hexapod Robot with Flexible Torso

Yiming Li, Bing Li

Robotic Intelligence

🎯 What it does: Designed and manufactured a hexapod robot with a flexible torso (F-RHex), and by observing the turning movements of Madagascar cockroaches, two bio-inspired turning strategies were extracted.

CoCMT: Communication-Efficient Cross-Modal Transformer for Collaborative Perception

Rujia Wang, Zhengzhong Tu

Autonomous DrivingComputational EfficiencyTransformerMultimodality

🎯 What it does: Proposed the CoCMT framework, which utilizes object queries to select and transmit key features, and constructed the Efficient Query Transformer (EQFormer) to efficiently fuse multi-agent object queries, combined with deep supervision to enhance overall performance.

CoCoL: A Communication Efficient Decentralized Collaborative Learning Method for Multi-Robot Systems

Jiaxing Huang, Jinming Xu

OptimizationFederated LearningComputational EfficiencyRobotic Intelligence

🎯 What it does: Propose the CoCoL method to achieve communication-efficient distributed collaborative learning

Code Generation and Monitoring for Deliberation Components in Autonomous Robots

Stefano Bernagozzi, Armando Tacchella

Robotic Intelligence

🎯 What it does: Propose modeling robot reasoning components at the conceptual level, automatically generating executable code and runtime monitors from the model, and subsequently validating their effectiveness in a simulated environment

Code-as-Symbolic-Planner: Foundation Model-Based Robot Planning via Symbolic Code Generation

Yongchao Chen, Chuchu Fan

OptimizationRobotic IntelligenceLarge Language Model

🎯 What it does: Utilizing large language models to generate code as symbolic planners in robot tasks and motion planning, addressing optimization and constraint verification problems.

CODE: COllaborative Visual-UWB SLAM for Online Large-Scale Metric DEnse Mapping

Lin Chen, Xuefeng Cao

Pose EstimationOptimizationSimultaneous Localization and MappingImage

🎯 What it does: Proposes a multi-UAV collaborative online dense mapping system that supports simultaneous collaborative localization and real-time dense map reconstruction, and recovers metric scale in environments without GNSS.

CoDifFu: Diffusion-Based Collaborative Perception with Efficient Heterogeneous Feature Fusion

ZeYu Meng, Jiayi Duan

Autonomous DrivingComputational EfficiencyDiffusion model

🎯 What it does: Propose the CoDifFu framework, leveraging a diffusion detection head and confidence-guided multi-body communication to achieve robust multi-agent perception

Coherent Online Road Topology Estimation and Reasoning with Standard-Definition Maps

Khanh Son Pham, C. Stachniss

Autonomous DrivingSimultaneous Localization and MappingImage

🎯 What it does: Propose a unified and coherent online HD map construction method, utilizing standard definition (SD) map information to predict lane segments, lane topology, and road boundaries, and enhancing network training stability and performance through hybrid lane segment encoding, denoising techniques, and temporal consistency from past frames.

ColaDex: Contact-guided Optimization and VLM-assisted Selection for Task-oriented Dexterous Grasp Generation

Yiyao Ma, Q. Dou

GenerationOptimizationRobotic IntelligencePrompt EngineeringVision Language Model

🎯 What it does: Proposed the ColaDex pipeline, which combines contact-based optimization to generate grasp candidates and utilizes a Vision-Language Model (VLM) for task-oriented grasp selection.

Collaborative Dynamic 3D Scene Graphs for Open-Vocabulary Urban Scene Understanding

Tim Steinke, Abhinav Valada

Autonomous DrivingGraph Neural NetworkWorld ModelImageMultimodalityPoint Cloud

🎯 What it does: Proposed the open-source vocabulary dynamic 3D scene graph engine CURB-OSG, which utilizes multi-agent collaboration to fuse camera and LiDAR observations, generating a hierarchical semantic structure for urban driving scenarios.

Collaborative Swarm Shape Reconstruction of Tumbling Space Targets via Decentralized Dynamic Factor Graph Optimization

El Ghali Asri, Zheng Hong Zhu

OptimizationSimultaneous Localization and Mapping

🎯 What it does: Developed a collaborative SLAM framework based on dynamic factor graphs for a drone fleet to collaboratively reconstruct the shape of an unknown rolling spatial object;

Collaborative Task Assignment, Sequencing and Multi-agent Path-finding

Yifan Bai, G. Nikolakopoulos

Optimization

🎯 What it does: Address the collaborative task allocation, scheduling, and multi-agent path planning problem, proposing the CBS-TS algorithm

Collective Behavior Clone with Visual Attention via Neural Interaction Graph Prediction

Kai Li, S. Zhao

Robotic IntelligenceGraph Neural NetworkReinforcement LearningAuto EncoderSequential

🎯 What it does: Proposes the Collective Behavior Cloning (CBC) framework, which uses a Graph Variational Autoencoder (GVAE) to learn local interaction graphs from group trajectories, and then learns group control strategies through behavior cloning based on the interaction graph and trajectories. Subsequently, it is deployed in real distributed visual robot swarms to achieve online neighbor selection.

Collective Motion of Magnetic Soft Cilia Controls Droplets and Nanozyme Catalysis*

Victoriia K. Lyashchuk, Daniil V. Kladko

Physics Related

🎯 What it does: Develop a magnetic soft bristle array and achieve high-frequency oscillation through an alternating magnetic field, enabling stable control of water droplets and enzyme-like catalytic reactions. Demonstrated water droplet transportation to a flexible pH sensor and a fourfold enhancement in nanoenzyme reaction rates.

Collision Avoidance with Differentiable Occupancy Functions in Object Rearrangement

Roma Satoh, Rei Kawakami

Robotic IntelligenceGaussian Splatting

🎯 What it does: By introducing a collision avoidance loss based on a differentiable occupancy function that considers object size, an additional training phase in existing models achieves a reduction in the number of collisions during transportation tasks while maintaining the semantic structure unchanged.

Collision Detection for Low-Cost Robot Manipulators Using Probabilistic Residual Torque Modeling

Y. Shao, Pengkang Yu

Representation LearningRobotic Intelligence

🎯 What it does: Propose a probability residual torque modeling method based on Gaussian Mixture Model (GMM) for collision detection in low-cost robotic manipulators.

Collision Mass Map for Safe and Efficient Human-Robot Interaction

Julian Balletshofer, Matthias Althoff

Safty and PrivacyData-Centric LearningRobotic Intelligence

🎯 What it does: Proposed and validated a data-based collision mass mapping for safety force prediction in human-robot interaction.

Collision-free Control Barrier Functions for General Ellipsoids via Separating Hyperplane

Zeming Wu, Lu Liu

Optimization

🎯 What it does: Proposed a general ellipsoid collision avoidance method based on Control Barrier Functions (CBF) and separating hyperplanes.

Collision-Free Trajectory Planning in Cluttered Environments for Efficient Bin Picking

Xiaomei Xu, Rainer Müller

Robotic IntelligencePoint Cloud

🎯 What it does: Developed a robotic arm trajectory planner for collision-free trajectory planning in cluttered reconfigurable environments, capable of quickly determining grasping paths and precisely assembling customized products.

CoMamba: Real-time Cooperative Perception Unlocked with State-Space Models

Jinlong Li, Zhengzhong Tu

Autonomous Driving

🎯 What it does: Propose the CoMamba framework, which realizes real-time vehicular collaborative 3D detection using state space models.

ComDrive: Comfort-Oriented End-to-End Autonomous Driving

Junming Wang, Wei Yin

Autonomous DrivingDiffusion model

🎯 What it does: Proposes ComDrive, an end-to-end autonomous driving system focused on comfort, for generating time-consistent and comfortable trajectories.

Compact LED-Based Displacement Sensing for Robot Fingers

Amr El-Azizi, M. Ciocarlie

Robotic Intelligence

🎯 What it does: Designed an LED-based displacement sensor for robotic fingers that can detect displacement caused by external forces.

Compact Modular Surgical System with a Novel RCM Mechanism for Laparo-Endoscopic Single-Site Surgery*

Z. Zhang, C.-K. Chui

Robotic Intelligence

🎯 What it does: A novel robotic-assisted single-port laparoscopic surgery system was proposed, with an arc-shaped host and a five-bar spherical parallel mechanism designed. An improved Denavit-Hartenberg parameterization was adopted to achieve remote center motion for multi-manipulators, followed by prototype experiments to verify its basic functions.

Compact R-X-Y Stage and Dual-Finger Micromanipulator under Inverted Optical Microscope for Microassembly

Jichao Pang, Xiaoming Liu

🎯 What it does: This study designs and implements a compact R-X-Y displacement platform and a two-finger micro manipulator, enabling long-distance X/Y displacement and 360° continuous rotation under an inverted microscope, and is used for microstructure assembly.

Comparative Analysis of CSP, CSTP, and Max-SNR Filters for P300 Detection in Brain Computer Interface

Saeid Piri, Huanghe Zhang

ClassificationBiomedical Data

🎯 What it does: Evaluate the performance of CSP, CSTP, and Max-SNR filtering methods in P300 detection, and propose a spatiotemporal filter based on Max-SNR

Competency-Aware Planning for Probabilistically Safe Navigation Under Perception Uncertainty

Sara Pohland, Claire J. Tomlin

Autonomous DrivingImage

🎯 What it does: Developed a probability and reconstruction-based capability estimation method called PaRCE, used to evaluate the familiarity of perception models with input images as a whole and their regions, and applied this capability information to navigation planning and control to ensure safe and efficient navigation under perception uncertainty.

Complete Corruption-Aware Retinex Framework for Low-Light Image Enhancement

Yifei Zhang, Kenji Hashimoto

RestorationImage

🎯 What it does: Propose a complete corruption-aware Retinex framework (CCRF), introducing a robust corruption-aware loss (RCL) and a Light-Up Map denoising (LMD) module for low-light image enhancement;

Complex Robotic Manipulation via Hindsight Goal Diffusion and Graph-based Experience Replay

Zi-Li Sun, Rui Song

Robotic IntelligenceReinforcement LearningDiffusion model

🎯 What it does: Proposed the HGD-GER method, achieving autonomous learning in robot manipulation tasks with obstacles and distant targets.

Compliant Tensegrity Robotic Arm with Continuously Adjustable Stiffness for Versatile Operation

David Herrmann, Valter Böhm

Robotic Intelligence

🎯 What it does: Designed an adjustable stiffness robotic arm based on a tensegrity structure, addressing the limitations of stiffness variation and cascading drive.

Composite Locally Weighted Learning Position and Stiffness Control of Articulated Soft Robots With Disturbance Observers

Zhigang Zou, Yongping Pan

Robotic Intelligence

🎯 What it does: Proposed a robust composite learning control (RCLC) scheme based on local weighted learning (LWL) for a multi-degree-of-freedom flexible robot (ASR) equipped with agonist-antagonist (AA) variable stiffness actuators (VSA), achieving good tracking of joint position and stiffness without requiring an exact robot model.

Computational Design of Closed Linkages For Robotic Limbs

Mikhail Chaikovskii, S. Kolyubin

OptimizationRobotic Intelligence

🎯 What it does: Proposed and implemented an open-source framework for optimizing the topology and parameters of closed-chain mechanical structures to achieve low inertia, high load capacity legged robots.

Computationally Efficient FPGA-based Large Language Model Inference for Real-Time Decision-Making in Robotic Systems

Huaizhi Zhang, Xiaojun Zhai

Computational EfficiencyRobotic IntelligenceTransformerLarge Language Model

🎯 What it does: Proposed a hardware optimization technique for deploying large language models on resource-constrained embedded devices, achieving significant reduction in computational latency through FPGA implementation;

Computing forward statics from tendon-length in flexible-joint hyper-redundant manipulators

Weiting Feng, Francesco Giorgio-Serchi

Robotic IntelligencePhysics Related

🎯 What it does: A method that unifies rope tension and length as inputs for forward statics solving was proposed, along with the development of a multi-segment ultra-redundant flexible joint cable-driven robotic arm model based on the screw formula. A forward statics iterative solving method was introduced, which can use either rope length or tension as input. Experimental validation demonstrated that the method works effectively with traditional tension input and when only rope length is used as input, achieving open-loop control using only kinematic inputs under static conditions, thereby avoiding practical challenges in tension measurement and state estimation.

Confidence-Controlled Exploration: Efficient Sparse-Reward Policy Learning for Robot Navigation

Kasun Weerakoon, Dinesh Manocha

Reinforcement Learning

🎯 What it does: Propose a confidence-based exploration method (CCE) that improves sample efficiency in sparse reward robot navigation by dynamically adjusting trajectory length according to policy entropy.

ConfigBot: Adaptive Resource Allocation for Robot Applications in Dynamic Environments

Rohit Dwivedula, Chris Rossbach

OptimizationRobotic Intelligence

🎯 What it does: Propose ConfigBot, which dynamically reconfigures robot applications to meet predefined performance specifications through runtime analysis and automatic configuration tuning, and experimentally validates its stability and resource optimization effects on multiple real robots.

Consensus-Driven Uncertainty for Robotic Grasping based on RGB Perception

Eric C. Joyce, Philippos Mordohai

Data SynthesisPose EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Train a lightweight deep network to predict the success of a pose estimation-guided grasp based on RGB images before execution

Consistent Feature Alignment for Cross-Modal Knowledge Distillation in Monocular 3D Object Detection

Fan Li, Xuguang Lan

Object DetectionAutonomous DrivingKnowledge DistillationImageMultimodalityPoint Cloud

🎯 What it does: Proposed a new network called MonoCFA, which achieves cross-modal knowledge distillation by designing a consistency alignment module and a deformable adapter module, thereby improving the performance of monocular 3D object detection.

Consistent Pose Estimation of Unmanned Ground Vehicles through Terrain-Aided Multi-Sensor Fusion on Geometric Manifolds

Alexander Raab, Abdalrahman Ibrahim

Pose EstimationAutonomous Driving

🎯 What it does: Proposes a pose estimation method for unmanned ground vehicles based on the Manifold Error State Extended Kalman Filter (M-ESEKF), aiming to improve the consistency and long-term accuracy of the Kalman filter;

Constrained Behavior Cloning for Robotic Learning

Jun Xie, Xiaoguang Ma

Robotic IntelligenceImageSequential

🎯 What it does: Proposes a geometric and historical constraint-based behavior cloning method (GHCBC), enhancing the learning performance of single-arm robots through visual and action history combined with high-level perceptual information.

ContactDexNet: Multi-fingered Robotic Hand Grasping in Cluttered Environments through Hand-Object Contact Semantic Mapping

Lei Zhang, Jianwei Zhang

GenerationRobotic IntelligenceAuto EncoderMultimodality

🎯 What it does: Propose the ContactDexNet method, which generates multi-finger grasping samples using hand-object contact semantic mapping, applicable in cluttered environments.

Contactless and Economical Chemical Reaction Platform Based on Ultrasonic Field

Yunsheng Li, Xiaoming Liu

Ultrasound

🎯 What it does: Established a contactless chemical reaction platform based on ultrasonic vortices, achieving stable capture, microdroplet transport, and multi-droplet mixing

Context-Aware Behavior Learning with Heuristic Motion Memory for Underwater Manipulation

Markus Buchholz, Yvan R. Pétillot

OptimizationRobotic Intelligence

🎯 What it does: Propose an adaptive heuristic motion planning framework that integrates heuristic motion space and Bayesian networks to enhance path planning performance in underwater manipulation.

Context-Aware Deep Lagrangian Networks for Model Predictive Control

L. Schulze, O. Arenz

OptimizationRobotic IntelligenceRecurrent Neural NetworkPhysics Related

🎯 What it does: Extend DeLaN to a context-aware model, integrate a recurrent network for online system identification, and combine with MPC to achieve adaptive physically consistent control

Context-Aware Multi-Agent Trajectory Transformer

Jeongho Park, Songhwai Oh

TransformerReinforcement LearningSequential

🎯 What it does: Proposed a Transformer-based multi-agent trajectory prediction model called COMAT for offline multi-agent reinforcement learning, which incorporates the historical information of neighboring agents as context for future trajectory prediction and applies it to planning

Context-aware Sparse Spatiotemporal Learning for Event-based Vision

Shenqi Wang, Guangzhi Tang

Object Detection

🎯 What it does: Propose the Context-aware Sparse Spatiotemporal Learning (CSSL) framework, which dynamically adjusts neuron activation through context-aware thresholds, applied to object detection and optical flow estimation in event cameras.

Context-Based Meta Reinforcement Learning for Robust and Adaptable Peg-in-Hole Assembly Tasks

Ahmed Shokry, Maren Bennewitz

Robotic IntelligenceMeta LearningReinforcement LearningImageTabular

🎯 What it does: Proposes a context-based meta-reinforcement learning method, training a PiH assembly agent using robot forward kinematics and uncalibrated camera information, while enhancing robustness and adaptability through efficient adaptation of force/torque sensor data and a process tailored for parameters beyond training tasks.

ContextCache: Task-Aware Lifecycle Management for Memory-Efficient LLM Agent Deployment

Tao Liu, Peng Wang

Computational EfficiencyLarge Language ModelAgentic AIBenchmark

🎯 What it does: Proposed a task-aware lifetime management framework called ContextCache to optimize context fragment caching for multi-step LLM agents, evaluated on a newly constructed dataset.

Continuous renal calculi tracking for autonomous robotic ureteroscopic lithotripsy

Quan Zhang, Yichao Tang

Object TrackingSegmentationRobotic IntelligenceConvolutional Neural NetworkReinforcement LearningImageBiomedical Data

🎯 What it does: Proposed an automated kidney stone fragmentation system integrating a robotic ureteroscope with sub-millimeter-level positioning accuracy, Quenching-net-based semantic segmentation, and collaborative deep reinforcement learning control.

Continuous-Time Gradient-Proportional-Integral Flow for Provably Convergent Motion Planning with Obstacle Avoidance

Jixiang Chen, Junzheng Wang

OptimizationOrdinary Differential Equation

🎯 What it does: Proposes a continuous-time gradient proportional integral flow (GPIF) for motion planning with obstacle avoidance, modeling the task as a constrained optimization problem relaxed into an unconstrained one, solving via gradient flow and function analysis, and enforcing constraint satisfaction through proportional and integral feedback.