arXivSub Start free trial

ICRA 2025 Papers — Page 3

IEEE International Conference on Robotics and Automation · 1604 papers

BiTrack: Bidirectional Offline 3D Multi-Object Tracking Using Camera-LiDAR Data

Kemiao Huang, Qi Hao

Object TrackingAutonomous DrivingMultimodalityPoint Cloud

🎯 What it does: Proposes the BiTrack framework, which includes modules such as 2D-3D detection fusion, initial trajectory generation, and bidirectional trajectory re-optimization, achieving offline 3D multi-object tracking for camera-radar data.

Blind Tactile Exploration for Surface Reconstruction

Yashaswi Sinha, Pradipta Biswas

Robotic IntelligenceMeshSequential

🎯 What it does: A method for blind tactile surface exploration using a series of sequential controllers is proposed for convex objects, achieving precise tactile exploration and high-accuracy surface reconstruction.

Blox-Net: Generative Design-for-Robot-Assembly Using VLM Supervision, Physics Simulation, and a Robot with Reset

A. Goldberg, Kenneth Y. Goldberg

GenerationRobotic IntelligenceVision Language ModelImageTextMultimodality

🎯 What it does: Proposes a system called Blox-Net that can generate robot-assemblable assembly plans and provide robot assembly instructions based on natural language prompts and images of available components.

BODex: Scalable and Efficient Robotic Dexterous Grasp Synthesis Using Bilevel Optimization

Jiayi Chen, He Wang

OptimizationRobotic IntelligenceBenchmark

🎯 What it does: Developed an efficient two-layer optimization grasp synthesis system and MuJoCo benchmark, generating a high-quality grasp dataset.

Boosting Cross-Spectral Unsupervised Domain Adaptation for Thermal Semantic Segmentation

Seokjun Kwon, Yukyung Choi

SegmentationDomain AdaptationImage

🎯 What it does: Research on cross-spectral unsupervised domain adaptation methods for thermal imaging semantic segmentation

Bootstrapping Object-Level Planning with Large Language Models

D. Paulius, G. Konidaris

Robotic IntelligenceLarge Language Model

🎯 What it does: Extract knowledge from large language models (LLMs) to generate object-level plans describing high-level state changes of objects, which guide task and motion planning (TAMP)

BoxMap: Efficient Structural Mapping and Navigation

Zili Wang, Roberto Tron

Autonomous DrivingRobotic IntelligenceTransformerSimultaneous Localization and Mapping

🎯 What it does: Proposes BoxMap, a structured mapping and navigation framework based on Detection-Transformer, which updates a topological map containing rooms, doors, and their connectivity relationships using partial environmental structures observed by sensors.

Braided Artificial Muscle with Programmable Body Morphing and its Application to Elbow Joint Flexion

Changchun Wu, Yonghua Chen

Robotic Intelligence

🎯 What it does: Propose a shape-deforming braided artificial muscle (SBAM), explain its basic structure, working mechanism, mathematical model, and validate its performance through experiments.

Brain-Inspired Spatial Continuous State Encoding for Efficient Spiking-Based Navigation

Qingao Chai, Huajin Tang

Autonomous DrivingSpiking Neural Network

🎯 What it does: Designed and implemented a navigation method based on spiking neural networks (SNNs), utilizing the population firing patterns of grid cells, head direction cells, and border vector cells to achieve continuous space state encoding, thereby enabling precise navigation target encoding and self-localization perception.

Bridging In-Situ and Satellite Data: Enhancing Gas Concentration Estimation Through Integration of Data-Driven and Physics-Based Modeling

Guoyu Lu

Data-Centric LearningPhysics Related

🎯 What it does: Integrate a physical model with a data-driven neural network framework, and refine a neural network model based on satellite observations using field measurement data to improve the estimation of trace gas concentrations such as NO2.

Bridging Spectral-Wise and Multi-Spectral Depth Estimation Via Geometry-Guided Contrastive Learning

Ukcheol Shin, Jean Oh

Depth EstimationContrastive LearningImage

🎯 What it does: This paper proposes a strategy named align-and-fuse for depth estimation from multispectral images. First, the embedding spaces of different spectral bands are aligned through geometry-guided contrastive learning. Then, a trainable attachable feature fusion module is introduced during the fusion stage to selectively aggregate multispectral features, achieving spectral-invariant and multispectral-fused depth estimation while maintaining reliability, memory efficiency, and flexibility.

Bridging the Human to Robot Dexterity Gap Through Object-Oriented Rewards

Irmak Güzey, Lerrel Pinto

Robotic IntelligenceReinforcement Learning from Human FeedbackSupervised Fine-TuningReinforcement LearningVideo

🎯 What it does: Construct a reward function generated from human videos using object-oriented rewards from a point tracker, online fine-tuning a multi-fingered robotic arm to perform tasks.

Bring Your Own Grasp Generator: Leveraging Robot Grasp Generation for Prosthetic Grasping

Giuseppe Stracquadanio, Lorenzo Natale

Pose EstimationRobotic Intelligence

🎯 What it does: Propose an eye-hand upper limb prosthetic grasping system that automatically configures the prosthetic hand's degrees of freedom and selects the grasp posture based on user intent

BUMBLE: Unifying Reasoning and Acting with Vision-Language Models for Building-wide Mobile Manipulation

Rutav Shah, Roberto Mart'in-Mart'in

Robotic IntelligenceTransformerVision-Language-Action ModelMultimodality

🎯 What it does: Proposes the BUMBLE framework, unifying visual language models with the perception, skills, and memory of long-range mobile manipulation;

C-Uniform Trajectory Sampling for Fast Motion Planning

O. G. Poyrazoglu, Volkan Isler

Autonomous DrivingOptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Study robot trajectory sampling, propose the concept of C-Uniform trajectories, provide a closed-form solution for 1D random walkers, utilize network flow optimization to precompute control action probabilities applicable to general robot systems, and apply this concept to the Model Predictive Path Integral controller, ultimately implementing experiments on a 1/10 scale race car.

CA-IoU: Central-Gaussian Angle-IoU for Robust Bounding Box Regression

Junbo Jang, Joonki Paik

Object Detection

🎯 What it does: Propose a new bounding box regression loss function called CA-IoU, which combines two novel loss terms: the central Gaussian integral and the angle IoU.

CaDRE: Controllable and Diverse Generation of Safety-Critical Driving Scenarios Using Real-World Trajectories

Peide Huang, Ding Zhao

GenerationAutonomous DrivingOptimizationSequential

🎯 What it does: Proposes a new framework called CaDRE for generating realistic, diverse, and controllable safety-critical driving scenarios.

CAFE-AD: Cross-Scenario Adaptive Feature Enhancement for Trajectory Planning in Autonomous Driving

Junrui Zhang, Yanyong Zhang

Autonomous DrivingPoint Cloud

🎯 What it does: Propose the CAFE-AD method, which enhances feature representation in trajectory planning through adaptive feature pruning and cross-scenario feature interpolation, addressing causal confusion and long-tail scenario distribution issues caused by open training.

CAFEs: Cable-Driven Collaborative Floating End-Effectors for Agriculture Applications

Hung Hon Cheng, Josie Hughes

Agriculture RelatedOrdinary Differential Equation

🎯 What it does: Developed a collaborative floating end-effector (CAFE) based on cable-driven systems to achieve automation for large-scale agricultural tasks.

Cage: Causal Attention Enables Data-Efficient Generalizable Robotic Manipulation

Shangning Xia, Cewu Lu

Robotic IntelligenceTransformerSupervised Fine-TuningDiffusion modelContrastive LearningImage

🎯 What it does: This paper proposes CAGE, a data-efficient and generalizable robotic manipulation strategy achieved by integrating pre-trained visual representations with causal attention mechanisms.

Camera-Lidar Consistent Neural Radiance Fields

Chao Hou, Fu Zhang

OptimizationNeural Radiance FieldImagePoint Cloud

🎯 What it does: Proposes a continuous-time framework for joint optimization of cameras and LiDAR, achieving more consistent radiance field reconstruction to enhance view synthesis and geometric accuracy.

Canonical Representation and Force-Based Pretraining of 3D Tactile for Dexterous Visuo-Tactile Policy Learning

Tianhao Wu, Hao Dong

Representation LearningRobotic Intelligence

🎯 What it does: Proposes a normalized representation for 3D tactile data and force-based self-supervised pre-training to enhance performance in multi-finger grasping tasks

CANVAS: Commonsense-Aware Navigation System for Intuitive Human-Robot Interaction

Suhwan Choi, Youngjae Yu

Robotic IntelligenceVision Language ModelVideoTextBenchmarkAgriculture Related

🎯 What it does: Proposes the CANVAS framework, integrating visual and language instructions for consensus-aware navigation, and employs imitation learning based on human navigation behavior.

CAO-RONet: A Robust 4D Radar Odometry with Exploring More Information from Low-Quality Points

Zhiheng Li, Zheng Fang

Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a learning-based 4D mmWave radar odometry framework, including local completion, context-aware hierarchical association, and window-based optimizer, to achieve robust ego-motion estimation under low-quality radar points.

Capacitated Agriculture Fleet Vehicle Routing with Implements and Limited Autonomy: A Model and a Two-Phase Solution Approach

Aitor López Sánchez, Holger Billhardt

OptimizationAgriculture Related

🎯 What it does: Proposes the Capacity-Constrained Agricultural Fleet Vehicle Routing Problem for Modular Autonomous Agricultural Robot Fleets (CAFVRPILA) and provides a two-phase heuristic decomposition solution.

CapBot: Enabling Battery-Free Swarm Robotics

Meng Liu, Danny Hughes

Robotic Intelligence

🎯 What it does: Developed a batteryless swarm robot named CapBot, utilizing supercapacitors to achieve fast charging, long-duration autonomous operation, and charge sharing.

Cascade IPG Observer for Underwater Robot State Estimation

Kaustubh Joshi, Nikhil Chopra

Robotic IntelligenceTime Series

🎯 What it does: Proposes a cascaded nonlinear observer framework for state estimation in underwater robots.

Cascaded Diffusion Models for Neural Motion Planning

Mohit Sharma, Oliver Kroemer

Autonomous DrivingDiffusion model

🎯 What it does: Proposes a cascading diffusion model for global motion planning, capable of generating complete trajectories and avoiding collisions through online planning repair.

CaStL: Constraints as Specifications Through Llm Translation for Long-Horizon Task and Motion Planning

Weihang Guo, L. Kavraki

Robotic IntelligenceLarge Language ModelText

🎯 What it does: The CaStL framework identifies constraints in natural language through multi-stage processing (such as goal conditions, action sequences, action blocking, etc.), leverages LLMs to translate these constraints into PDDL and Python scripts, and solves task and motion planning problems using a custom PDDL solver.

Catch It! Learning to Catch in Flight with Mobile Dexterous Hands

Yuanhang Zhang, Huazhe Xu

Robotic IntelligenceReinforcement Learning

🎯 What it does: A mobile manipulator consisting of a mobile base, a 6-DoF robotic arm, and a 12-DoF dexterous hand was constructed, along with training a whole-body control strategy to enable the robot to capture flying objects.

Categorical Traffic Transformer: Interpretable and Diverse Behavior Prediction with Tokenized Latent

Yuxiao Chen, Marco Pavone

Autonomous DrivingExplainability and InterpretabilityTransformer

🎯 What it does: Proposed a traffic model that simultaneously outputs continuous trajectory prediction and semantic classification prediction

Causal Contrastive Learning with Data Augmentations for Imitation-Based Planning

Haojie Xin, Z. Yang

Autonomous DrivingContrastive LearningBenchmark

🎯 What it does: Proposed the \overline{C}^{2}L framework, combining contrastive learning with counterfactual data augmentation, and introduced a shortcut eliminator to extract non-redundant features and reduce temporal spurious correlations.

CC-STAR: An Estimation for Contact State Transition Using Reconstruction-Based Anomaly Detection for Peg-in-Hole Assembly

Haeseong Lee, Jaeheung Park

Anomaly DetectionConvolutional Neural NetworkAuto EncoderTime Series

🎯 What it does: Proposed an anomaly detection framework based on input data reconstruction to estimate contact state transitions in plug-in assembly.

CDM: Contact Diffusion Model for Multi-Contact Point Localization

Seohee Han, Min Jun Kim

Robotic IntelligenceDiffusion model

🎯 What it does: Proposes a multi-contact point localization method CDM based on diffusion models, designed for robots equipped with joint torque sensors and base force/torque sensors to perform multi-contact point localization.

CDMFusion: RGB-T Image Fusion Based on Conditional Diffusion Models via Few Denoising Steps in Open Environments

Luojie Yang, Yufeng Yue

RestorationDiffusion modelMultimodality

🎯 What it does: Proposes CDMFusion, a three-branch conditional diffusion model for RGB-T image fusion, which adaptively enhances multimodal features through a dynamic gating module and accelerates the generation process using a skip patrol mechanism.

CELLmap: Enhancing LiDAR SLAM Through Elastic and Lightweight Spherical Map Representation

Yifan Duan, Yanyong Zhang

Autonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Designed an elastic lightweight spherical map representation called CELLmap, and proposed a general backend based on CELL's bidirectional registration module and loop closure detection module

Certificated Actor-Critic: Hierarchical Reinforcement Learning with Control Barrier Functions for Safe Navigation

Junjun Xie, Huijun Gao

Autonomous DrivingReinforcement Learning

🎯 What it does: Proposed and implemented a model-free reinforcement learning algorithm called Certificated Actor-Critic (CAC), which employs a hierarchical reinforcement learning framework and utilizes an explicit reward function based on control barrier functions.

CGTrack: Cascade Gating Network with Hierarchical Feature Aggregation for UAV Tracking

Weihong Li, Libo Zhang

Object Tracking

🎯 What it does: Proposes CGTrack, a UAV tracker based on Hierarchical Feature Cascade (HFC) and Lightweight Gated Center Head (LGCH), which expands network capacity through a coarse-to-fine two-tier framework.

Chain-of-Modality: Learning Manipulation Programs from Multimodal Human Videos with Vision-Language-Models

Chen Wang, Jacky Liang

Robotic IntelligencePrompt EngineeringVision Language ModelVideoMultimodalityChain-of-ThoughtAudio

🎯 What it does: Integrate multimodal data such as video, electromyography (EMG) signals, and audio into visual language models using Chain-of-Modality (CoM), extracting task plans and generating fine-grained control parameters, enabling robots to perform the same operational tasks based on a single multimodal human video prompt.

Chameleon: Fast-Slow Neuro-Symbolic Lane Topology Extraction

Zongzheng Zhang, Hao Zhao

Autonomous DrivingComputational EfficiencyVision Language ModelImageChain-of-Thought

🎯 What it does: Proposed a fast-slow neuro-symbolic lane topology extraction algorithm called Chameleon, which alternates between using a fast system for direct inference of detection instances and a slow system that leverages VLM chain-of-thought to handle edge cases

Chemistry3D: Robotic Interaction Toolkit for Chemistry Experiments

Shoujie Li, Wenbo Ding

Robotic IntelligenceReinforcement LearningImage

🎯 What it does: Proposed the Chemistry3D robot interaction toolkit, enabling robots to perform chemical experiments and real-time visualization of temperature, color, and pH changes.

CLIMB: Language-Guided Continual Learning for Task Planning with Iterative Model Building

Walker Byrnes, Animesh Garg

Robotic IntelligenceMeta LearningReinforcement Learning from Human FeedbackLarge Language ModelTextSequential

🎯 What it does: Propose the CLIMB framework, which utilizes a base model and execution feedback for language-guided continual learning to iteratively build a domain model for robot task planning; the framework can construct models from natural language descriptions, learn hidden predicates, and store information for future use.

Closed Loop Interactive Embodied Reasoning for Robot Manipulation

Michal Nazarczuk, K. Mikolajczyk

Robotic IntelligenceVision-Language-Action ModelBenchmark

🎯 What it does: A modular closed-loop interactive embodied reasoning (CLIER) method is proposed for robot manipulation, considering non-visual attribute measurements, scene changes caused by external disturbances, and uncertain outcomes of robot actions. It performs multi-modal reasoning and action planning to generate executable primitive action sequences.

Closed-Loop Open-Vocabulary Mobile Manipulation with GPT-4V

Peiyuan Zhi, Siyuan Huang

Robotic IntelligenceLarge Language ModelVision Language Model

🎯 What it does: Proposes a closed-loop mobile and desktop manipulation robot system called COME-robot, which utilizes GPT-4V to achieve open-vocabulary perception and situational reasoning, and completes task verification and failure recovery through an iterative closed-loop feedback mechanism.

CloudTrack: Scalable UAV Tracking with Cloud Semantics

Yannik Blei, Wolfram Burgard

Object TrackingVision Language ModelImageText

🎯 What it does: Proposes a scalable drone-based object tracking method leveraging cloud semantics, enabling open-vocabulary tracking based on verbal descriptions (e.g., shirt color).

CLSTR: Capability-Level System for Tracking Robots

Alexandra Bejarano, Tom Williams

Robotic Intelligence

🎯 What it does: Proposed the CLSTR system to visualize the capabilities and member changes of dynamic robot teams, helping human operators maintain situational awareness.

CNSv2: Probabilistic Correspondence Encoded Neural Image Servo

An-Jen Chen, Yue Wang

Robotic IntelligenceImage

🎯 What it does: Developed and implemented CNSv2, a visual servoing method that utilizes probabilistic feature matching to enhance robustness and accuracy in scenarios with inconsistent lighting or missing textures.

Co-MTP: A Cooperative Trajectory Prediction Framework with Multi-Temporal Fusion for Autonomous Driving

Xinyu Zhang, Hong Chen

Autonomous DrivingGraph Neural NetworkTransformerSequential

🎯 What it does: Proposes Co-MTP, a collaborative trajectory prediction framework that utilizes multi-temporal fusion to capture historical and future interaction information using V2X technology.

Coarse-to-Fine Cross-Modality Generation for Enhancing Vehicle Re-Identification with High-Fidelity Synthetic Data

Leyang Jin, Zhedong Zheng

RecognitionData SynthesisAutonomous DrivingPrompt EngineeringDiffusion modelImageBenchmark

🎯 What it does: This paper proposes a coarse-to-fine generation pipeline for synthesizing high-fidelity vehicle data, supporting the training and learning of vehicle re-identification models.

Cocube: a Tabletop Modular Multi-Robot Platform for Education and Research

Shuai Liang, Xuelong Li

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed the CoCube desktop modular multi-robot platform, aimed at supporting robot education and multi-robot algorithm research.

CoDynTrust: Robust Asynchronous Collaborative Perception via Dynamic Feature Trust Modulus

Yunjiang Xu, Jianping Wang

Object DetectionAutonomous Driving

🎯 What it does: Proposes the CoDynTrust framework, leveraging dynamic feature trust modulus (DFTM) and a multi-scale fusion module to achieve robust fusion for asynchronous collaborative perception.

CognitiveOS: Large Multimodal Model Based System to Endow Any Type of Robot with Generative AI

Artem Lykov, D. Tsetserukou

GenerationRobotic IntelligenceTransformerAgentic AIVision-Language-Action ModelMultimodality

🎯 What it does: Built a cognitive robot operating system called CognitiveOS that can run on multiple robot platforms, utilizing a multimodal large model multi-agent system, modular and configurable, supporting complex tasks.

Cohere3D: Exploiting Temporal Coherence for Unsupervised Representation Learning of Vision-Based Autonomous Driving

Yichen Xie, Xin Huang

Autonomous DrivingRepresentation LearningContrastive LearningImagePoint Cloud

🎯 What it does: Proposed a contrastive learning algorithm called Cohere3D, which learns instance representations in vision-based Bird's-eye-view (BEV) feature maps under unsupervised conditions, maintaining consistency over long time sequences.

COHERENT: Collaboration of Heterogeneous Multi-Robot System with Large Language Models

Kehui Liu, Xuelong Li

Robotic IntelligenceLarge Language ModelTextBenchmark

🎯 What it does: Proposes the COHERENT framework, which utilizes large language models (LLMs) for task planning in heterogeneous multi-robot collaboration, and designs a Proposal-Execution-Feedback-Adjustment (PEFA) loop mechanism.

COIGAN: Controllable Object Inpainting Through Generative Adversarial Network for Defect Synthesis in Data Augmentation

Massimiliano Biancucci, P. Zingaretti

GenerationData SynthesisGenerative Adversarial NetworkImage

🎯 What it does: Proposed a controllable object inpainting generative adversarial network (COIGAN) to synthesize realistic defect images for expanding the predictive maintenance dataset.

CoL3D: Collaborative Learning of Single-view Depth and Camera Intrinsics for Metric 3D Shape Recovery

Chenghao Zhang, Jieping Ye

Depth EstimationImagePoint Cloud

🎯 What it does: Propose the CoL3D framework, which employs a unified network to achieve cooperative learning for monocular depth estimation and camera intrinsic parameter estimation, aiming to recover metric 3D shapes.

COLA: Characterizing and Optimizing the Tail Latency for Safe Level-4 Autonomous Vehicle Systems

Haolan Liu, Jishen Zhao

Autonomous DrivingOptimization

🎯 What it does: Systematic study on the causes of tail latency in autonomous vehicle (AV) systems and its impact on safety

Collaborative Motion Planning for Multi-Manipulator Systems Through Reinforcement Learning and Dynamic Movement Primitives

Siddharth Singh, Qing Chang

Robotic IntelligenceReinforcement Learning

🎯 What it does: A multi-level collaborative motion planning method is proposed, integrating reinforcement learning and dynamic movement primitives to generate adaptive real-time trajectories for multi-robotic arm systems, achieving collision avoidance and improved collaborative efficiency.

COLLAGE: Collaborative Human-Agent Interaction Generation Using Hierarchical Latent Diffusion and Language Models

Divyanshu Daiya, Aniket Bera

GenerationData SynthesisTransformerLarge Language ModelPrompt EngineeringDiffusion modelAuto EncoderVideoText

🎯 What it does: Propose the COLLAGE framework, using LLM and hierarchical VQ-VAE to generate collaborative human-object-human interactions

Collapsible Airfoil Single Actuator ROtor-Craft (CASARO) - Construction and Analysis of a Soft Rotary Wing Robot

W. Ang, S. Foong

Robotic Intelligence

🎯 What it does: Designed, constructed, controlled, and flight tested the foldable flexible single-motor rotor robot CASARO.

Collision Avoidance in Model Predictive Control Using Velocity Damper

Arthur Haffemayer, Nicolas Mansard

OptimizationRobotic Intelligence

🎯 What it does: Proposed a model predictive control method based on velocity damper constraints to achieve strict collision avoidance for robotic arms in dynamic environments;

Collision-Aware Traversability Analysis for Autonomous Vehicles in the Context of Agricultural Robotics

F. Philippe, Jean-Philippe Lauffenburger

Autonomous DrivingImagePoint CloudAgriculture RelatedPhysics Related

🎯 What it does: This paper proposes a feasibility analysis method for collision perception based on 3D spectrum maps constructed using LiDAR and multispectral cameras, aimed at safe navigation of agricultural robots. The method evaluates the safety of robot movement on flexible obstacles through physics-based feasibility metrics.

Combined Modal Robust Cascade Control for Wheeled Self-Reconfigurable Robots Under Drive Failure and Safety Threat

Tao Jiang, Yizhuo Sun

Robotic Intelligence

🎯 What it does: Based on nonholonomic constraints and the Lagrangian method, a WSRR (Wheeled-Segway Robotic Rower) combination modal kinematics and dynamics model with arbitrary reconfigurable scale was constructed; a smooth obstacle avoidance strategy based on a safety fence was designed; an adaptive fault-tolerant mechanism was introduced to reasonably distribute torque and avoid degradation of tracking performance; the Extended State Observer (IESO) was improved to suppress high-frequency oscillations of measurement noise and initial observation error peaks, achieving robust speed tracking control under unknown aggregated disturbances; finally, the fault tolerance, robustness, and safety of the method were verified on a real WSRR experimental platform.

Command Filtered Cartesian Impedance Control for Tendon Driven Continuum Manipulators with Actuator Fault Compensation

Xianjie Zheng, Yu Guo

Robotic Intelligence

🎯 What it does: This paper proposes a finite-time phase Cartesian impedance control scheme for tension-driven continuous deformation manipulators (TDCM), combining command filtered backstepping with a second-order low-pass filter to achieve real-time reference trajectory adjustment of external end-effector forces. Additionally, a dedicated actuator fault compensation algorithm is designed to address failures in tension transmission, and experimental validation of trajectory tracking is conducted on multi-segment TDCM prototypes.

Commonsense Reasoning for Legged Robot Adaptation with Vision-Language Models

Annie S. Chen, Chelsea Finn

Robotic IntelligenceVision Language ModelMultimodality

🎯 What it does: Proposed the VLM-Predictive Control (VLM-PC) system, which uses vision-language models to enable real-time adaptive behavior selection for quadruped robots, helping them navigate complex obstacle environments.

Communication-Aware Iterative Map Compression for Online Path-Planning

Evangelos Psomiadis, Panagiotis Tsiotras

CompressionAutonomous DrivingOptimizationImage

🎯 What it does: Proposed a task-driven, communication-aware map compression framework that supports mobile robots in heterogeneous resource-constrained teams by using compressed maps to assist navigation

Comparison of Three Interface Approaches for Gaze Control of Assistive Robots for Individuals with Tetraplegia

E. N. Sardinha, Marcela Múnera

Robotic IntelligenceBiomedical DataBenchmark

🎯 What it does: This study evaluated the performance and user experience of three control interfaces (GUI, embedded interface, directional gaze) in assistive robotic arms.

Comparison of User Interface Paradigms for Assistive Robotic Manipulators

Amelia Sinclaire, H. Yanco

Robotic Intelligence

🎯 What it does: A within-subjects study conducted with 27 participants over the age of 60, comparing two interaction methods on an assistive skateboard robot: a graphical user interface (GUI) using a 10-inch touchscreen versus a physical user interface incorporating a joystick, button box, and projector.

Complementary Information Guided Occupancy Prediction via Multi-Level Representation Fusion

Rongtao Xu, Xiaodan Liang

Autonomous DrivingKnowledge DistillationRepresentation LearningConvolutional Neural NetworkGraph Neural NetworkPoint CloudBenchmark

🎯 What it does: Proposes the CIGOcc two-stage occupancy prediction framework, integrating segmentation, graphical, and depth features, and introduces a deformable multi-layer fusion mechanism and SAM knowledge distillation.

Compliance Control with Dynamic and Self-Sensing Hydraulic Artificial Muscles for Wearable Assistive Devices

Bibhu Sharma, Thanh Nho Do

Robotic Intelligence

🎯 What it does: Proposed a dynamic and self-sensing control scheme leveraging the inherent soft sensing capabilities of fluidic thread actuators, enabling soft wearable robots to achieve simultaneous sensing and actuation with 96% position accuracy, 70.5% tracking accuracy, and a response delay of only 0.09 seconds on a flexible assistive arm.

Component-Aware Unsupervised Logical Anomaly Generation for Industrial Anomaly Detection

Xuan Tong, Wenqiang Zhang

GenerationData SynthesisAnomaly DetectionImage

🎯 What it does: Proposes an unsupervised, component-aware anomaly generation framework named ComGEN, aiming to generate logically consistent industrial anomaly images.

Composing Dextrous Grasping and In-Hand Manipulation via Scoring with a Reinforcement Learning Critic

Lennart Röstel, Berthold Bauml

Robotic IntelligenceReinforcement LearningScore-based Model

🎯 What it does: Utilizing a reinforcement learning critic network to score and select initial grasps, thereby integrating grasping with in-hand manipulation.

Composite Learning Neural Network Tracking Control of Articulated Soft Robots

Zhigang Zou, Yongping Pan

Robotic Intelligence

🎯 What it does: Developed a combination learning controller based on neural networks for position tracking control of joint soft robots driven by positive and negative antagonistic variable stiffness actuators.

Computational Teaching for Driving via Multi-Task Imitation Learning

D. Gopinath, Avinash Balachandran

Autonomous DrivingTime SeriesSequential

🎯 What it does: We propose an automated high-performance driving instruction system trained through multi-task imitation learning (MTIL), aiming to simulate the interactions between human instructors and learners;

Computationally and Sample Efficient Safe Reinforcement Learning Using Adaptive Conformal Prediction

Hao Zhou, Wenhao Luo

Computational EfficiencyReinforcement Learning

🎯 What it does: Proposed a sample-efficient, periodic safety learning framework for online control tasks, integrating Gaussian process kernel approximation, adaptive conformal prediction, control barrier functions, and optimistic exploration strategies.

ConceptAgent: LLM-Driven Precondition Grounding and Tree Search for Robust Task Planning and Execution

Corban Rivera, R. Chellappa

Robotic IntelligenceTransformerLarge Language ModelMultimodality

🎯 What it does: Propose the ConceptAgent platform, combining LLM-driven precondition induction and tree search to enable task planning and execution for robots in open-world environments.

Configuration-Adaptive Visual Relative Localization for Spherical Modular Self-Reconfigurable Robots

Yuming Liu, T. Lam

Object TrackingPose EstimationOptimizationRobotic IntelligenceImage

🎯 What it does: Proposed and implemented a configurable adaptive visual relative positioning method for spherical modular self-reconfigurable robots (SMSR).

Conformalized Reachable Sets for Obstacle Avoidance with Spheres

Yong-Seog Kwon, Ram Vasudevan

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Proposed CROWS, a real-time, recursive window trajectory planner that generates probabilistically safe motion planning;

Constrained Learning for Decentralized Multi-Objective Coverage Control

J. Cerviño, A. Ribeiro

OptimizationRobotic Intelligence

🎯 What it does: Proposed a decentralized constraint learning method combining primal-dual optimization with a learnable perception-action-communication neural network (LPAC) for multi-target coverage control.

Constrained Nonlinear Kaczmarz Projection on Intersections of Manifolds for Coordinated Multi-Robot Mobile Manipulation

Akshaya Agrawal, Geoffrey A. Hollinger

Robotic Intelligence

🎯 What it does: Propose modeling various constraints in multi-robot collaborative manipulation as a manifold set, generating solutions that satisfy constraints using restricted nonlinear Kaczmarz projection techniques, and subsequently combining them with sampling-based motion planning algorithms to produce complex coordinated motions of 3–6 mobile manipulators (18–36 degrees of freedom).

Contact Force Estimation for a Leg-Wheel Transformable Robot With Varying Contact Points

Yi-Syuan Shen, Pei-Chun Lin

Robotic IntelligenceTime Series

🎯 What it does: Proposes a new technology for contact force estimation in deformable leg-wheel robots, capable of evaluating contact forces across the entire leg-wheel surface (including areas such as the tip and sides);

Context Graph-Based Visual-Language Place Recognition

Soojin Woo, Seong-Woo Kim

RetrievalRepresentation LearningConvolutional Neural NetworkVision Language ModelMultimodality

🎯 What it does: Proposed a visual-language place recognition method based on a zero-shot language-driven semantic segmentation model, leveraging pixel-level embeddings to construct semantic image descriptors;

Context-Aware Collaborative Pushing of Heavy Objects Using Skeleton-Based Intention Prediction

Gokhan Solak, Arash Ajoudani

Pose EstimationGraph Neural Network

🎯 What it does: The study investigates cooperative pushing and pulling tasks involving heavy objects on frictional surfaces, using a pose-based context-aware method to predict human motion intent.

Contingency Formation Planning for Interactive Drone Light Shows

T. Au

Robotic IntelligenceVideo

🎯 What it does: The study utilizes drone formations as a video game platform, combining planning techniques to achieve accurate display of animated pixels and rapid response to user input.

Continuous Control of Diverse Skills in Quadruped Robots Without Complete Expert Datasets

Jiaxin Tu, Lihua Zhang

Robotic IntelligenceGenerative Adversarial Network

🎯 What it does: Propose a new quadruped robot multi-skill continuous control method called PASIST without requiring a complete expert dataset, which autonomously explores and selects high-quality trajectories using target poses, and achieves skill transfer through a generative adversarial self-imitation learning framework.

Continuous Convolution for Automated Measurement of Sperm Flagella

Yufei Jin, Zhuoran Zhang

SegmentationConvolutional Neural NetworkBiomedical Data

🎯 What it does: Developed an automatic high-throughput tool based on continuous convolution for quantitative analysis of sperm flagellar vibrations.

Continuous Wrist Control on the Hannes Prosthesis: A Vision-Based Shared Autonomy Framework

F. Vasile (Italian Institute of Technology), Lorenzo Natale (Italian Institute of Technology)

Robotic Intelligence

🎯 What it does: Using a computer vision-based shared autonomy framework for continuous motion control of a prosthetic wrist, achieving more natural grasping paths.

ConTrack3D: Contrastive Learning Contributes Concise 3D Multi-Object Tracking

Ruibin Du, Rui Feng

Object TrackingContrastive LearningPoint CloudBenchmark

🎯 What it does: Propose ConTrack3D, an online end-to-end joint detection and tracking system

Contrastive Touch-to-Touch Pretraining

Samanta Rodriguez, Nima Fazeli

Representation LearningContrastive LearningMultimodality

🎯 What it does: This paper learns a unified representation that captures shared information between different tactile sensors (GelSlim and Soft Bubble) and maps their touch signals to a shared embedding space.

Control Reallocation Using Deep Reinforcement Learning for Actuator Fault Recovery of an Autonomous Underwater Vehicle

Katell Lagattu, Benoît Clement

Robotic IntelligenceReinforcement Learning

🎯 What it does: Using deep reinforcement learning for control reallocation to achieve actuator fault recovery in unmanned underwater vehicles

Control Strategies for Pursuit-Evasion Under Occlusion Using Visibility and Safety Barrier Functions

Minnan Zhou, Nikolay Atanasov

Autonomous DrivingOptimization

🎯 What it does: Developed a control strategy to maintain the visibility of the target within the pursuer's field of view in occluded environments, achieving pursuit and obstacle avoidance through visibility and safety control barrier functions.

Controlled Robot Language with Frame Semantics (FrameCRL) for Autonomous Context-Aware High-Level Planning

Dang M. Tran, Hongsheng He

Robotic IntelligenceLarge Language ModelText

🎯 What it does: Propose a configurable and scalable planning generation framework based on controlled robot language and frame semantics (FrameCRL); generate discourse representation structures (DRS) from natural language instructions, extract imperative verbs, align with the FrameNet framework to construct goal declarations in planning language, further analyze non-imperative sentences to generate object specifications and initial states, and merge all statements into a planning script directly solvable by an integrated planner; evaluate on multiple natural language corpora and demonstrate robust language understanding in dual-arm robot grasping-placing tasks.

CoopDETR: A Unified Cooperative Perception Framework for 3D Detection via Object Query

Zhe Wang, Ya-Qin Zhang

Object DetectionAutonomous DrivingTransformerPoint Cloud

🎯 What it does: Proposes a unified collaborative perception framework named CoopDETR, using object queries to achieve object-level feature collaboration for 3D detection.

Cooperative Distributed Model Predictive Control for Embedded Systems: Experiments with Hovercraft Formations

Gösta Stomberg, T. Faulwasser

OptimizationPhysics Related

🎯 What it does: Implement embedded distributed model predictive control on a set of suspended gliders on an air hockey table, and conduct formation control experiments in various scenarios.

Coordinated Multi-Robot Navigation with Formation Adaptation

Zihao Deng, Hao Zhang

Robotic IntelligenceGraph Neural NetworkReinforcement Learning

🎯 What it does: Proposes the Adaptive Formation with Oscillation Reduction (AFOR) method for achieving multi-robot collaborative navigation and formation adaptation, utilizing hierarchical learning, spring-damper model, graph neural networks (GNN), and reinforcement learning (RL) to realize team coordination and individual control.

coVoxSLAM: GPU Accelerated Globally Consistent Dense SLAM

Emiliano Höss, Pablo De Crist'oforis

Simultaneous Localization and Mapping

🎯 What it does: Proposed and implemented a GPU-accelerated global consistent dense SLAM system called coVoxSLAM

CRAB: Camera-Radar Fusion for Reducing Depth Ambiguity in Backward Projection Based View Transformation

I. Lee, Dongsuk Kum

Object DetectionSegmentationDepth EstimationAutonomous DrivingImagePoint Cloud

🎯 What it does: Proposed a camera-radar fusion-based 3D object detection and segmentation model named CRAB, which employs inverse projection combined with radar information to reduce depth ambiguity.

Cross-Platform Learning-Based Fault Tolerant Surfacing Controller for Underwater Robots

Yuya Hamamatsu, A. Ristolainen

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a cross-platform fault-tolerant buoyancy controller based on reinforcement learning, which enables underwater robots to achieve buoyancy control using only the remaining functional actuators without explicitly identifying the failed actuators;

CrowdSurfer: Sampling Optimization Augmented with Vector-Quantized Variational AutoEncoder for Dense Crowd Navigation

Naman Kumar, K. Krishna

Autonomous DrivingOptimizationAuto Encoder

🎯 What it does: Proposes CrowdSurfer, which learns expert trajectory priors through a Vector-Quantized Variational AutoEncoder (VQ-VAE) and uses it as initialization for sampling optimization during runtime, thereby improving local planning efficiency in crowded environments.

CtRNet-X: Camera-to-Robot Pose Estimation in Real-World Conditions using a Single Camera

Jingpei Lu, Michael C. Yip

Pose EstimationRobotic IntelligenceVision Language ModelImage

🎯 What it does: Proposes a framework that can estimate robot pose when the robot is partially visible.