IROS 2024 Papers — Page 4
IEEE/RSJ International Conference on Intelligent Robots and Systems · 1581 papers
Combining Sampling- and Gradient-based Planning for Contact-rich Manipulation
Filippo Rozzi, Kevin Haninger
OptimizationRobotic Intelligence
🎯 What it does: Propose a planning method that combines sampling and gradient optimization, using the Cross-Entropy Method to initialize the gradient solver and integrating a particle filter in the sampling-based planner for online contact mode estimation.
Commonsense Scene Graph-based Target Localization for Object Search
Wenqi Ge, Hong Zhang
Robotic IntelligenceGraph Neural NetworkLarge Language ModelPoint CloudGraph
🎯 What it does: Propose a target localization method based on a common-sense scene graph, CSG-TL, to enhance target search for home robots in dynamic environments.
Communication-Constrained Multi-Robot Exploration with Intermittent Rendezvous
Alysson Ribeiro Da Silva, Ani Hsieh
OptimizationRobotic Intelligence
🎯 What it does: Propose modeling the multi-robot exploration task as a Distributed Partially Observable Markov Decision Process (DEC-POMDP), and design a joint strategy following a rendezvous plan, enabling robots to occasionally share maps in unknown environments without being constrained by joint path optimization.
CompdVision: Combining Near-Field 3D Visual and Tactile Sensing Using a Compact Compound-Eye Imaging System
Lifan Luo, Hongyu Yu
Depth EstimationRobotic IntelligenceMultimodality
🎯 What it does: Designed and implemented the CompdVision sensor, a compound eye imaging system that integrates near-field 3D vision and tactile perception.
Competitive Multi-Team Behavior in Dynamic Flight Scenarios
T. Seyde, Daniela Rus
Reinforcement Learning
🎯 What it does: Propose a hierarchical control model that combines high-level strategic guidance with low-level task-specific continuous control to achieve collaborative coordination among competing teams in dynamic aerial scenarios (F-16 aircraft)
Compliance Optimization Control for Rigid-Soft Hybrid System and its Application in Humanoid Robot Motion Control
Zewen He, Ko Yamamoto
OptimizationRobotic Intelligence
🎯 What it does: A control framework based on compliance optimization for a rigid-soft hybrid robot system is proposed, utilizing a piecewise constant strain (PCS) model to describe the continuous deformation of flexible structures. The system is divided into two states: single support and double support, and the method is validated through forward dynamics simulation on a human-like robot with a flexible prosthetic limb.
Compliant Blind Handover Control for Human-Robot Collaboration
Davide Ferrari, Cristian Secchi
Robotic Intelligence
🎯 What it does: Proposes a human-robot blind handover architecture, enabling the robot to autonomously complete the entire handover process in blind handover scenarios, with a focus on safety and release timing detection.
ComTraQ-MPC: Meta-Trained DQN-MPC Integration for Trajectory Tracking with Limited Active Localization Updates
Gokul Puthumanaillam, M. Ornik
Autonomous DrivingOptimizationMeta LearningReinforcement Learning
🎯 What it does: Proposed and implemented the ComTraQ-MPC framework, integrating a meta-trained DQN with MPC to achieve trajectory tracking under conditions where active localization updates are limited.
Conditional Generative Denoiser for Nighttime UAV Tracking
Yucheng Wang, Haobo Zuo
Object TrackingTransformerVideo
🎯 What it does: Proposed a conditional generative denoiser (CG-Denoiser) for nighttime UAV target tracking, breaking the deterministic paradigm by conditionally generating and removing noise on the input.
Conditional Variational Autoencoders for Probabilistic Pose Regression
Fereidoon Zangeneh, P. Jensfelt
Pose EstimationAuto EncoderImage
🎯 What it does: Propose a probabilistic method based on conditional variational autoencoders (CVAEs) for predicting the posterior distribution of camera pose given an image, and build a generative model through training strategies that can produce this posterior distribution, enabling sampling of multiple possible poses.
Confidence-Aware Decision-Making and Control for Tool Selection*
A. Meera, Pablo Lanillos
OptimizationRobotic Intelligence
🎯 What it does: Proposed a mathematical framework for control confidence calculation and decision-making in dynamic systems, and applied it to the tool selection problem;
Consistent Distributed Cooperative Localization: A Coordinate Transformation Approach
Chungeng Tian, Haodi Yao
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposes a distributed collaborative localization algorithm for multi-robot systems that performs state estimation under coordinate transformations to address consistency issues.
Constrained 6-DoF Grasp Generation on Complex Shapes for Improved Dual-Arm Manipulation
Gaurav Singh, K. M. Krishna
GenerationRobotic IntelligenceDiffusion model
🎯 What it does: Proposes CGDF, which utilizes diffusion models to generate 6-DoF grasp poses on specific regions of complex shapes to support dual-arm operations.
Constrained Bootstrapped Learning for Few-Shot Robot Skill Adaptation
Nadim Haque, Teresa Vidal-Calleja
Robotic IntelligenceMeta LearningReinforcement Learning
🎯 What it does: Propose a hybrid learning method combining demonstration learning and reinforcement learning for robot skill learning, capable of online rapid adaptation to new tasks.
Construction of Musculoskeletal Simulation for Shoulder Complex with Ligaments and Its Validation via Model Predictive Control
Yuta Sahara, Masayuki Inaba
OptimizationBiomedical Data
🎯 What it does: A detailed shoulder joint simulation model incorporating the scapula, clavicle, joint ligaments, and maximum muscle strength was constructed, and the stabilizing role of ligaments during initial motion and the load-balancing effect of muscle force distribution were validated through model predictive control simulations.
Contact Stability Control of Stepping Over Partial Footholds Using Plantar Tactile Feedback
J. R. Guadarrama-Olvera, Gordon Cheng
Robotic Intelligence
🎯 What it does: Using plantar tactile feedback to detect terrain geometry and online reconstruct the support polygon after each step landing; calculating the convex hull from contact points detected by distributed forward force sensors, then using the polygon centroid for repositioning ZMP and DCM positions, with the support polygon serving as a constraint for ZMP balance feedback control.
Contact-Implicit Model Predictive Control for Dexterous In-hand Manipulation: A Long-Horizon and Robust Approach
Yongpeng Jiang, Xiang Li
OptimizationRobotic Intelligence
🎯 What it does: Proposed a long-horizon dexterous grasping control method based on a high-level contact-implicit model predictive controller.
ContactHandover: Contact-Guided Robot-to-Human Object Handover
Zixing Wang, Shuran Song
Pose EstimationOptimizationRobotic Intelligence
🎯 What it does: Proposes a robot-to-human object handover system called ContactHandover, consisting of a contact-guided grasping phase and an object delivery phase.
Contacts from Motion: Learning Discrete Features for Automatic Contact Detection and Estimation from Human Movements *
H. Miyake, Eiichi Yoshida
OptimizationTransformerAuto EncoderTime Series
🎯 What it does: Solely utilizing human motion data, automatically detect and estimate contact forces and contact positions using machine learning methods
Context-Aware Conversation Adaptation for Human-Robot Interaction
Zhidong Su, Weihua Sheng
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed and implemented a Context-Aware Dialogue Adaptation System (CACAS) based on environmental context, enhancing the proactivity and intelligence of human-computer interaction through context recognition, language processing, and strategies combining reinforcement learning with neural networks.
Context-Aware GAN-based Image Retrieval for Coarse Localization of Autonomous Robots
Ruphan Swaminathan, Pradyot V. N. Korupolu
Pose EstimationRetrievalAutonomous DrivingGenerative Adversarial NetworkImagePoint Cloud
🎯 What it does: Proposed a context-based GAN (ConLocGAN) that extracts robust global features for coarse localization of autonomous delivery robots through a two-step process combining image retrieval and LiDAR localization.
Context-Generative Default Policy for Bounded Rational Agent
Durgakant Pushp, Lantao Liu
OptimizationRobotic IntelligenceDiffusion model
🎯 What it does: Proposed a context-generated default policy enabling robots to predict unobserved environments using observed regions, thereby dynamically adjusting decisions
Contextual Emotion Recognition using Large Vision Language Models
Yasaman Etesam, Angelica Lim
RecognitionTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelMultimodality
🎯 What it does: Explored two methods for emotion recognition using large vision-language models, evaluated on the EMOTIC dataset.
Continual Domain Randomization
Josip Josifovski, Nicolás Navarro-Guerrero
Domain AdaptationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a Continual Domain Randomization (CDR) framework that combines domain randomization with continual learning to achieve more robust simulation-to-reality transfer in reinforcement learning by sequentially training subsets of randomized parameters in simulation.
Continual Learning for Autonomous Robots: A Prototype-based Approach
Elvin Hajizada, Yulia Sandamirskaya
Robotic IntelligenceMeta LearningSpiking Neural NetworkImage
🎯 What it does: Proposed and implemented a prototype-based continuous learning method CLP, capable of learning in few-shot online continuous learning scenarios and detecting and learning novel objects without supervision.
Continuous Rapid Learning by Human Imitation using Audio Prompts and One-Shot Learning
Jaime Duque Domingo, J. García-Bermejo
Robotic IntelligenceMeta LearningPrompt EngineeringAudio
🎯 What it does: Propose a flexible neural learning architecture that enables robots to learn to grasp objects by merely observing human demonstrations and quickly adapt and execute tasks based on simple voice prompts from users.
Continuum Robot Shape Estimation Using Magnetic Ball Chains
G. Pittiglio, P. Dupont
Robotic IntelligenceBiomedical Data
🎯 What it does: Proposed and verified a method for real-time shape sensing of medical continuum robots using magnetic ball chains and Hall effect sensor arrays.
Contrastive Mask Denoising Transformer for 3D Instance Segmentation
He Wang, Guofeng Zhang
SegmentationTransformerContrastive LearningPoint Cloud
🎯 What it does: Proposed a contrastive mask denoising Transformer, consisting of a mask denoising module and a multi-modal perceptual query selection module, for 3D instance segmentation.
Control of Unknown Quadrotors from a Single Throw
Till M. Blaha, B. Remes
Robotic Intelligence
🎯 What it does: This paper proposes a method to restore a quadrotor drone after being thrown without knowing any control parameters. It utilizes high-frequency rotational speed feedback to estimate control effectiveness and fit the motor model using recursive least squares estimation. An excitation sequence is designed within the gyroscope measurement limits to provide large attitude control. Subsequently, a 52-parameter incremental nonlinear dynamic inversion (INDI) attitude controller is used to stop rotation and restore vertical attitude within 450 ms. Then, a nonlinear dynamic inversion (NDI) position controller drives the drone to a target position. The algorithm can efficiently run on common microcontrollers and successfully restores the quadrotor in experiments every time, even when thrown from a height of only 3.5 meters, and is robust to initial rotation and noise. In simulations, the root mean square error of parameter fitting for randomly thrown quadrotors is typically within 10%.
Control-Oriented Reinforcement Active Modeling Scheme for Hysteresis Compensation of Flexible Endoscopic Robot
Fan Ren, Jianda Han
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed an extended Kalman filter (EKF) active modeling approach combined with reinforcement learning for lag compensation in flexible endoscopic robots.
CoNVOI: Context-aware Navigation using Vision Language Models in Outdoor and Indoor Environments
A. Sathyamoorthy, Dinesh Manocha
Robotic IntelligenceTransformerPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Propose the CoNVOI method, which utilizes Vision Language Models (VLM) for context-aware navigation in indoor and outdoor environments, generating human-like movement paths.
Cooperative Modular Manipulation with Numerous Cable-Driven Robots for Assistive Construction and Gap Crossing
Kevin F. Murphy, João Ramos
Robotic Intelligence
🎯 What it does: Developed a distributed multi-module cable-driven robot system named Co3MaNDR, capable of achieving precise trajectory tracking and force amplification by simultaneously pulling a central load through multiple modules. Hardware experiments have demonstrated remote load operation, trajectory following, and operator force perception and amplification.
Cooperative Path Planning for Four-Way Shuttle Vehicles in Storage and Retrieval Systems: A Hierarchically Dynamic Graph-Based Approach
Xingyao Han, Hesheng Wang
OptimizationGraph
🎯 What it does: A collaborative path planning method with dynamic adjustment of road map structure is proposed for large-scale collaborative operations of four-way shuttle cars in storage and retrieval systems.
Coordinated Multi-arm 3D Printing using Reeb Decomposition
Jayant Khatkar, Ramgopal R. Mettu
OptimizationRobotic IntelligenceMeshBenchmark
🎯 What it does: Proposes a collaborative multi-arm 3D printing framework using Reeb decomposition to enable collaborative manufacturing by multiple extruders in a shared workspace.
CoPa: General Robotic Manipulation through Spatial Constraints of Parts with Foundation Models
Haoxu Huang, Yang Gao
Robotic IntelligenceVision Language Model
🎯 What it does: Propose the CoPa framework, achieving robot manipulation based on component spatial constraints through a base model
CoT-TL: Low-Resource Temporal Knowledge Representation of Planning Instructions Using Chain-of-Thought Reasoning
Kumar Manas, A. Paschke
Explainability and InterpretabilityRepresentation LearningTextChain-of-Thought
🎯 What it does: Propose the CoT-TL framework for converting natural language planning instructions into linear temporal logic (LTL) representations in low-resource scenarios.
CR3DT: Camera-RADAR Fusion for 3D Detection and Tracking
N. Baumann (ETH Zurich), Michele Magno (ETH Zurich)
Object DetectionObject TrackingAutonomous DrivingImageMultimodality
🎯 What it does: Proposed CR3DT, a 3D object detection and multi-object tracking model that fuses camera and radar data.
Creating Discomfort Maps via Hand-held Human Feedback Interface for Robotic Shoulder Physiotherapy
Jevon Ravenberg, Luka Peternel
Robotic IntelligenceReinforcement Learning from Human Feedback
🎯 What it does: Proposed a method to capture patients' discomfort during robotic shoulder physical therapy through a handheld human-computer feedback interface, creating a 'discomfort map'.
Cross-Architecture Auxiliary Feature Space Translation for Efficient Few-Shot Personalized Object Detection
Francesco Barbato, Mete Ozay
Object DetectionKnowledge DistillationTransformerContrastive LearningImage
🎯 What it does: Propose a few-shot personalized object detection framework AuXFT, integrating coarse-to-fine hierarchical learners and a feature translator to achieve cross-architecture feature mapping.
Cross-Modal Self-Supervised Learning with Effective Contrastive Units for LiDAR Point Clouds
Mu Cai, Xiaodong Yang
Object DetectionSegmentationAutonomous DrivingContrastive LearningPoint Cloud
🎯 What it does: Research and propose a cross-modal self-supervised contrastive learning method, as well as an instance-aware and similarity-balanced contrastive unit tailored for LiDAR, and conduct experiments on multiple benchmarks.
Cross-Modal Visual Relocalization in Prior LiDAR Maps Utilizing Intensity Textures
Qiyuan Shen, Ming Yang
Pose EstimationRetrievalSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Proposes a cross-modal visual localization system based on intensity texture, consisting of three modules: map projection, coarse retrieval, and fine localization.
Cross-Observability Learning for Vehicle Routing Problems
Ruifan Liu, Antonios Tsourdos
OptimizationReinforcement Learning
🎯 What it does: To address the limited observation range in vehicle routing problems, a multi-agent learning framework based on cross-observability is proposed, along with a proof that the theoretical upper bound of the optimality gap for imitation strategies decreases as the neighborhood range expands; subsequently, the MACOPO algorithm is introduced, which solves multi-vehicle routing problems under limited observation by leveraging fully observed experts to guide cross-entropy optimization.
Crowd-Aware Robot Navigation with Switching Between Learning-Based and Rule-Based Methods Using Normalizing Flows
Kohei Matsumoto, Ryo Kurazume
Robotic IntelligenceReinforcement LearningFlow-based Model
🎯 What it does: Propose a switching method for mobile robot navigation in crowded environments, which can dynamically select learning-based or rule-based control strategies according to the scene.
CRPlace: Camera-Radar Fusion with BEV Representation for Place Recognition
Shaowei Fu, Yanyong Zhang
RetrievalAutonomous DrivingTransformerImagePoint Cloud
🎯 What it does: Proposed a camera-radar fusion-based background attention method CRPlace for scene recognition
CSR: A Lightweight Crowdsourced Road Structure Reconstruction System for Autonomous Driving
Huayou Wang, Changliang Xue
Autonomous DrivingOptimizationComputational EfficiencySimultaneous Localization and Mapping
🎯 What it does: Propose a lightweight crowdsourcing-based road structure reconstruction system CSR, which achieves vectorized reconstruction of road structures relying solely on online-perceived semantic elements
CTS: Sim-to-Real Unsupervised Domain Adaptation on 3D Detection
Meiying Zhang, Qi Hao
Object DetectionDomain AdaptationPoint Cloud
🎯 What it does: Transfer the model trained on labeled simulated data to unlabeled real data to achieve unsupervised domain adaptation for 3D detection.
CubiX: Portable Wire-Driven Parallel Robot Connecting to and Utilizing the Environment
Shintaro Inoue, Masayuki Inaba
Robotic Intelligence
🎯 What it does: Developed a portable parallel robot named CubiX that achieves interaction with the environment through the winding of multiple cables
Current-Based Impedance Control for Interacting with Mobile Manipulators
Jelmer de Wolde, Javier Alonso-Mora
Robotic Intelligence
🎯 What it does: Proposed a current-based impedance control method to achieve compliance control on mobile robots without force/torque sensors, and verified its performance on the Kinova GEN3 Lite robotic arm.
CurricularVPR: Curricular Contrastive Loss for Visual Place Recognition
Dongshuo Zhang, S. Lam
RecognitionContrastive Learning
🎯 What it does: Proposed a Curricular Contrastive Loss (CCL) and utilized graded similarity labels to measure sample difficulty, thereby improving the training efficiency and effectiveness of visual place recognition.
D-MARL: A Dynamic Communication-Based Action Space Enhancement for Multi Agent Reinforcement Learning Exploration of Large Scale Unknown Environments
Gabriele Calzolari, G. Nikolakopoulos
Reinforcement Learning
🎯 What it does: Propose a dynamic communication-based action space enhancement method for D-MARL exploration algorithms to improve mapping efficiency in unknown environments (occupancy grid maps).
D2SR: Decentralized Detection, De-Synchronization, and Recovery of LiDAR Interference
Darshana Rathnayake, Archan Misra
Depth EstimationAnomaly DetectionAutonomous DrivingComputational EfficiencyPoint Cloud
🎯 What it does: Proposed the D2SR method, which achieves detection removal, desynchronization, and depth recovery for multi-LiDAR interference.
D3G: Learning Multi-robot Coordination from Demonstrations
Yizhi Zhou, Xuan Wang
OptimizationRobotic Intelligence
🎯 What it does: Developed a distributed method for solving the inverse problem of differentiable dynamic games (D3G), enabling robots to learn multi-robot collaboration from demonstrations.
DaDiff: Domain-aware Diffusion Model for Nighttime UAV Tracking
Haobo Zuo, Jia Pan
Object TrackingDomain AdaptationDiffusion modelVideoBenchmark
🎯 What it does: Proposed a stepwise alignment paradigm named DaDiff to align low-resolution target features in nighttime UAV tracking to daytime features, and constructed the NUT-LR nighttime UAV tracking benchmark dataset.
DailySTR: A Daily Human Activity Pattern Recognition Dataset for Spatio-temporal Reasoning
Yue Qiu, Ryusuke Sagawa
RecognitionData SynthesisTransformerLarge Language ModelVideoBenchmark
🎯 What it does: Proposes a synthetic video question answering dataset called DailySTR that focuses on the spatiotemporal relationships of daily human activities, and introduces a two-stage model that first uses Transformer to decode video content and then employs a large language model (LLM) for cross-video spatiotemporal reasoning.
DAP: Diffusion-based Affordance Prediction for Multi-modality Storage
Haonan Chang, Abdeslam Boularias
Pose EstimationRobotic IntelligenceDiffusion modelMultimodalityBenchmark
🎯 What it does: Proposes a diffusion-based affordance prediction (DAP) process to address multi-modal object storage problems, first locating placeable regions and then precisely calculating relative poses.
DarkGS: Learning Neural Illumination and 3D Gaussians Relighting for Robotic Exploration in the Dark
Tianyi Zhang, Matthew Johnson-Roberson
Robotic IntelligenceGaussian Splatting
🎯 What it does: Construct a re-lightable 3D high-fidelity scene representation under low illumination and moving light source environments, and assist scene reconstruction through a learned illumination model.
Data Efficient Behavior Cloning for Fine Manipulation via Continuity-based Corrective Labels
Abhay Deshpande, S. Srinivasa
Data-Centric LearningRobotic IntelligenceWorld Model
🎯 What it does: Studied the effectiveness of the Continuous Continuity-based Correction Label (CCIL) framework in real-world fine-grained manipulation tasks for alleviating covariate shift (compounding errors).
Data-driven Force Observer for Human-Robot Interaction with Series Elastic Actuators using Gaussian Processes
Samuel Tesfazgi, Sandra Hirche
Robotic Intelligence
🎯 What it does: A data-driven dynamic observer based on Gaussian process regression was developed for estimating interaction forces in physical human-robot interaction under series elastic actuators (SEA), while considering model uncertainty.
Data-Driven Koopman Operator-Based Error-State Kalman Filter for Enhanced State Estimation of Quadrotors in Agile Flight
Peng Huang, Gerhard Fettweis
Robotic IntelligenceTime SeriesPhysics Related
🎯 What it does: Propose a data-driven attitude estimation method based on the Koopman operator and error-state Kalman filter (K-ESKF), applicable to high dynamic flight.
Data-Driven Modeling of Cable Slab Dynamics via Neural Networks
Y. Al-Rawashdeh, M. Janaideh
Object TrackingPose EstimationVideo
🎯 What it does: Using a trained neural network to track the positions of multiple markers on the cable plate from real-time visual feedback captured by a high-speed camera, and generating time-varying cubic B-spline curves to approximate the two-dimensional dynamics and bending geometry of the cable plate.
Data-Driven Modeling of Ground Effect For UAV Landing on a Vertical Oscillating Platform
Binglin He, Yang Wang
OptimizationRobotic Intelligence
🎯 What it does: Construct a data-driven ground effect dynamic model and integrate it with a feedforward controller to address the landing problem of multirotor drones on vertically oscillating platforms.
Data-Driven Predictive Control for Robust Exoskeleton Locomotion
Kejun Li, A. D. Ames
Robotic Intelligence
🎯 What it does: Proposes a data-driven predictive control (DDPC) method for lower-limb exoskeleton gait generation, constructing a model-free predictive control framework using the Hankel matrix and state transition matrix
Data-Driven System Identification of Quadrotors Subject to Motor Delays
Jonas Eschmann, Giuseppe Loianno
Robotic IntelligenceTime SeriesPhysics Related
🎯 What it does: A data-driven method is used to identify the inertial parameters, thrust curves, torque coefficients, and first-order motor delays of a quadrotor based solely on ontology perception data;
DCSANet: Dual Cross-channel and Spatial Attention Make RGB-T Object Detection Better
Xiaoxiong Lan, Changzhen Qiu
Object DetectionImageMultimodality
🎯 What it does: Propose a lightweight feature enhancement fusion module (FEM), composed of a channel enhancement fusion unit (CEU) and a spatial enhancement fusion unit (SEU), to improve the feature fusion effectiveness in RGB-T object detection.
DD-VNB: A Depth-based Dual-Loop Framework for Real-time Visually Navigated Bronchoscopy
Qingyao Tian, Hongbin Liu
Pose EstimationDepth EstimationSimultaneous Localization and MappingBiomedical Data
🎯 What it does: Proposed a depth-based dual-loop framework DD-VNB for real-time visual navigation bronchoscope localization.
DDS-SLAM: Dense Semantic Neural SLAM for Deformable Endoscopic Scenes
Jiwei Shan, Hesheng Wang
Simultaneous Localization and MappingBiomedical Data
🎯 What it does: Propose DDS-SLAM for achieving accurate camera tracking, continuous dense reconstruction, and high-quality image rendering in deformable endoscopic scenarios.
DEAR: Disentangled Environment and Agent Representations for Reinforcement Learning without Reconstruction
Ameya Pore, D. Dall'Alba
Representation LearningReinforcement LearningImage
🎯 What it does: Proposes a disentangled environment and agent representation method called DEAR for vision-based reinforcement learning using proxy segmentation masks without reconstruction;
DECADE: Towards Designing Efficient-yet-Accurate Distance Estimation Modules for Collision Avoidance in Mobile Advanced Driver Assistance Systems
M. Shahzad, Muhammad Shafique
Depth EstimationAutonomous DrivingConvolutional Neural NetworkSupervised Fine-TuningImagePoint Cloud
🎯 What it does: Designed and implemented the DECADE model for fast distance estimation in mobile ADAS, utilizing detector outputs and pose estimation DNN to predict object distance.
DECAF: a Discrete-Event based Collaborative Human-Robot Framework for Furniture Assembly
Giulio Giacomuzzo, Diego Romeres
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a discrete-event collaborative human-robot framework called DECAF for furniture assembly, utilizing discrete-event Markov decision processes (DE-MDP) and reinforcement learning to achieve optimal action planning for robots when collaborating with humans.
DecAP : Decaying Action Priors for Accelerated Imitation Learning of Torque-Based Legged Locomotion Policies
Shivam Sood, Guillaume Sartoretti
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a two-phase framework that first generates imitation data by training a position control policy, and then improves the exploration and learning efficiency of the torque control policy using decaying action priors;
Decentralized Acceleration-Based Bird-Inspired Flocking
Luca Iacone, Dario Albani
Robotic Intelligence
🎯 What it does: Designed and implemented a decentralized group control method based on acceleration for UAV formation movement
Decentralized Collaborative Localization and Map Update with Buildings
Maxime Escourrou, P. Bonnifait
Autonomous DrivingSimultaneous Localization and MappingPoint Cloud
🎯 What it does: Achieve decentralized collaborative localization and map updating in urban environments using building façade features detected by 3D LiDAR, leveraging direct communication between vehicles.
Decentralized Communication-Maintained Coordination for Multi-Robot Exploration: Achieving Connectivity and Adaptability
W. Tang, Qiuguo Zhu
Robotic IntelligenceGraph Neural NetworkTransformer
🎯 What it does: Propose a decentralized multi-robot exploration algorithm that ensures the robot swarm maintains continuous communication.
Decentralized Linear Convoying for Underactuated Surface Craft with Partial State Coupling
Raymond Turrisi, Michael Benjamin
Autonomous Driving
🎯 What it does: Proposed a decentralized linear formation algorithm and control method, implemented on the MOOS-IvP platform; achieved platform independence through an abstract layer virtual system; discretized the leader's trajectory and embedded it into the leader's dynamics, propagating it to followers; employed a PD controller to eliminate cumulative errors and enhance tracking performance; further strengthened formation cohesion through virtual coupling of partial states.
Decentralized Multi-Robot Navigation Coupled with Spatial-Temporal RetNet Based on Deep Reinforcement Learning
Lin Chen, Danwei Wang
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a multi-robot navigation policy network named Spatial-Temporal RetNet (STR), and validated its collision avoidance and task completion performance in dynamic multi-robot scenarios within simulation environments.
Decentralized Trajectory Planning for Formation Flight in Unknown and Dense Environments
Jianxin Zeng, Hesheng Wang
Autonomous DrivingOptimization
🎯 What it does: Propose a decentralized trajectory planning framework to achieve formation flight in unknown dense environments
Deep Ad-hoc Sub-Team Partition Learning for Multi-Agent Air Combat Cooperation
Songyuan Fan, Roushu Yang
Graph Neural NetworkReinforcement LearningGraph
🎯 What it does: Proposed Deep Ad-hoc Sub-Team Partition Learning (DASPL), which models multi-agent aerial combat as a graph, automatically partitions dynamic sub-teams, transforms large-scale cooperative problems into multiple small-scale equivalent problems, and introduces efficient inter-subteam information passing.
Deep Domain Adaptation Regression for Force Calibration of Optical Tactile Sensors
Zhuo Chen, Shan Luo
Domain Adaptation
🎯 What it does: Propose an unsupervised force calibration method based on deep domain adaptation, transferring the force prediction capability of calibrated optical tactile sensors to uncalibrated sensors, addressing domain differences in labeling, illumination, elastic modulus, etc.
Deep Geometric Potential Functions for Tracking on Manifolds
N. Prakash, R. Horowitz
Object Tracking
🎯 What it does: Proposed a potential function defined by a neural network that is invariant on a manifold, used to generate elastic forces for achieving asymptotic trajectory tracking.
Deep Learning-based Delay Compensation Framework For Teleoperated Wheeled Rovers on Soft Terrains
Ahmad Abubakar, Lakmal Seneviratne
Robotic IntelligenceRecurrent Neural Network
🎯 What it does: Developed a model-free predictive framework based on deep learning to compensate for network latency in remotely operated wheeled rovers on soft terrain, enhancing command tracking performance.
Deep Sensor Fusion with Constraint Safety Bounds for High Precision Localization
Sebastian Schmidt, Stephan Günnemann
Safty and PrivacyRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposed a novel data-driven but provably bounded sensor fusion method and applied it to mobile robot localization.
Deep Stochastic Kinematic Models for Probabilistic Motion Forecasting in Traffic
L. Zheng, Ming C. Lin
Autonomous DrivingStochastic Differential Equation
🎯 What it does: By predicting actions and performing time integration using a kinematic bicycle model, probabilistically predict the vehicle's future trajectory, and describe the variance relationships between time steps using a differentiable analytical approximation.
Deep Visual Odometry with Events and Frames
R. Pellerito, Davide Scaramuzza
Autonomous DrivingRecurrent Neural NetworkSimultaneous Localization and MappingMultimodality
🎯 What it does: Proposed and implemented RAMP-VO, an end-to-end learning image and event camera fusion visual odometry system
DeepBHMR: Learning Bidirectional Hybrid Mixture Models for Generalized Rigid Point Set Registration
Zhe Min, M. Q. Meng
Pose EstimationPoint CloudBiomedical Data
🎯 What it does: Propose a deep learning-based rigid registration method called DeepBHMR, which achieves precise registration by utilizing normals and a bidirectional hybrid model.
Deeper Introspective SLAM: How to Avoid Tracking Failures Over Longer Routes?
Kanwal Naveed, Donghwan Lee
TransformerReinforcement LearningSimultaneous Localization and MappingVideo
🎯 What it does: Reveals the limitations of depth reinforcement learning-based visual SLAM in long-term trajectory tracking, and proposes a new architecture based on video vision Transformers to enhance the long-term tracking capability, successfully completing longer trajectories.
DeepMIF: Deep Monotonic Implicit Fields for Large-Scale LiDAR 3D Mapping
Kutay Yilmaz, A. Artemov
Autonomous DrivingOptimizationNeural Radiance FieldSimultaneous Localization and MappingPoint CloudBenchmark
🎯 What it does: Propose a LiDAR large-scale 3D mapping method based on deep monotonic implicit fields, optimizing non-metric monotonic implicit fields to achieve dense 3D scenes.
Deformable Objects Perception is Just a Few Clicks Away – Dense Annotations from Sparse Inputs
Alessio Caporali, Gianluca Palli
SegmentationRobotic IntelligenceImage
🎯 What it does: Proposes a method to generate pixel-level dense segmentation labels for flexible objects from sparse keypoint annotations, combined with images collected by a robot camera.
Demonstrating a Robust Walking Algorithm for Underactuated Bipedal Robots in Non-flat, Non-stationary Environments
Oluwami Dosunmu-Ogunbi, J. Grizzle
Robotic Intelligence
🎯 What it does: Propose a robust algorithm for underactuated bipedal robots to walk in non-flat, non-static environments, combining ankle torque and an improved angular momentum linear inverted pendulum (ALIP) model, and achieving gait stability through a dual-strategy controller (virtual constraints + ALIP-MPC); hardware deployment and demonstration of mobile path walking functionality on the Cassie robot.
Demonstrating Trustworthiness in Open-Loop Model Mediated Teleoperation for Collecting Lunar Regolith Simulant*
Joe Louca, R. Charles
Computational EfficiencyRobotic IntelligencePhysics Related
🎯 What it does: Implemented a computationally efficient lunar regolith model and compared it with real physical equivalents in an open-loop model-mediated teleoperation system.
Demonstration to Adaptation: A User-Guided Framework for Sequential and Real-Time Planning
Kuanqi Cai, Sami Haddadin
OptimizationRobotic IntelligenceSequential
🎯 What it does: This paper proposes a user-guided planning framework for dynamic human-centered environments, enabling robots to encode object-related constraints and user preferences through multiple demonstrations, transfer geometric features and implicit relaxation to new scenarios, and adapt in real-time to task changes.
Density-aware Domain Generalization for LiDAR Semantic Segmentation
Jaeyeul Kim, Sunghoon Im
SegmentationDomain AdaptationConvolutional Neural NetworkPoint Cloud
🎯 What it does: Proposes two methods that enable the network to perform adaptive operations based on point cloud density without any constraints, thereby maintaining consistent performance across different density environments.
Depth Completion using Galerkin Attention
Yinuo Xu, Xuesong Zhang
Depth EstimationConvolutional Neural NetworkImage
🎯 What it does: Densify sparse depth maps using the Galerkin attention mechanism.
Depth Helps: Improving Pre-trained RGB-based Policy with Depth Information Injection
Xincheng Pang, Xuelong Li
Depth EstimationSupervised Fine-TuningReinforcement LearningImage
🎯 What it does: Proposed the Depth Information Injection framework, which fine-tunes pre-trained RGB policies using depth information and deploys using only RGB images during inference.
DeRO: Dead Reckoning Based on Radar Odometry With Accelerometers Aided for Robot Localization
H. Do, J. Song
Robotic IntelligenceSimultaneous Localization and MappingMultimodalityTime Series
🎯 What it does: Proposed a radar odometry structure that directly computes pose using 4D FMCW radar Doppler velocity and gyroscope data, with state updates performed within a Kalman filter framework.
DESectBot: Design and Validation of a Novel Two-Segment Decoupled Continuum Robotic System for Endoscopic Submucosal Dissection
Wenjie Liu, Peng Qi
Robotic Intelligence
🎯 What it does: Designed and verified a two-stage decoupled continuum robot system named DESectBot for endoscopic submucosal dissection (ESD) surgery.
Design and Control of a Novel Six-Degree-of-Freedom Hybrid Robotic Arm
Yang Chen, Ya Xiong
Robotic IntelligenceAgriculture Related
🎯 What it does: A new six-degree-of-freedom hybrid robotic arm called LingXtend was proposed, combining parallel and serial mechanisms. It achieves obstacle avoidance and a significant expansion of the workspace through two independent slides and a gear system.
Design and Control of a Novel Soft-Rigid Lower Limb Exoskeleton Robot
Yuxuan Wang, Yanqiong Fei
Robotic IntelligenceBiomedical Data
🎯 What it does: Proposed a novel soft-rigid lower limb exoskeleton robot, completed its structural design, CPAM characteristic modeling, hybrid feedforward-feedback control, and EMG-based human-robot interaction control, and verified its performance through experiments.
Design and Control of a Soft Supernumerary Robotic Limb Based on Fiber-Reinforced Actuator
Tianyi Zhang, Youfu Li
Robotic Intelligence
🎯 What it does: Designed and implemented a soft redundant robotic arm based on fiber-reinforced actuators (FRA), established its kinematic model and control system, and subsequently validated the accuracy of the control strategy through experiments.
Design and Control of a Three-Dimensional Electromagnetic Drive System for Micro-Robots
Yunrui Zhang, Qigao Fan
OptimizationRobotic IntelligencePhysics Related
🎯 What it does: Designed and implemented a three-dimensional electromagnetic drive system, and conducted driving and path tracking experiments on micro-robots.
Design and Control of an Ultra-Slender Push-Pull Multisection Continuum Manipulator for In-Situ Inspection of Aeroengine
Weiheng Zhong, Nianfeng Shao
Robotic Intelligence
🎯 What it does: Developed and verified an ultra-slim push-pull multi-segment continuum manipulator for in-situ inspection of multi-stage blades in aircraft engines.