arXivSub Start free trial

IROS 2024 Papers — Page 6

IEEE/RSJ International Conference on Intelligent Robots and Systems · 1581 papers

Efficient Incremental Penetration Depth Estimation between Convex Geometries

Wei Gao

Optimization

🎯 What it does: Proposes an optimization-based incremental penetration depth estimation algorithm capable of calculating the minimum penetration depth and its direction between convex geometries.

Efficient Motion Prediction: A Lightweight & Accurate Trajectory Prediction Model With Fast Training and Inference Speed

Alexander Prutsch, Horst Possegger

Autonomous DrivingOptimizationComputational Efficiency

🎯 What it does: Proposed a lightweight, short-training-time, and fast-inference motion prediction model

Efficient Multimodal Semantic Segmentation via Dual-Prompt Learning

Shaohua Dong, Heng Fan

SegmentationComputational EfficiencyPrompt EngineeringMultimodality

🎯 What it does: Proposed DPLNet, which adapts frozen pre-trained RGB models to multi-modal semantic segmentation through dual prompt learning, achieving efficient fusion with only a small number of trainable parameters.

Efficient Path Planning for Modular Reconfigurable Robots

M. Mayer, Matthias Althoff

OptimizationRobotic IntelligenceBenchmark

🎯 What it does: Studied accelerating path planning in modular reconfigurable robots by migrating existing paths to new assemblies;

Efficient Tactile Sensing-based Learning from Limited Real-world Demonstrations for Dual-arm Fine Pinch-Grasp Skills

Xiaofeng Mao, Zhibin Li

Robotic IntelligenceConvolutional Neural NetworkAuto EncoderMultimodality

🎯 What it does: Proposed a dual-arm fine-grained grasping control framework that enables data-efficient learning using rich tactile perception data.

Efficient Target Singulation with Multi-fingered Gripper using Propositional Logic

Hyojeong Kim, ChangHwan Kim

OptimizationRobotic Intelligence

🎯 What it does: Propose a search method based on propositional logic for target singulation of densely packed utensils using a multi-fingered gripper, ensuring that the generated relocation plans have global optimality and can quickly determine if singulation is impossible.

Efficient Trajectory Forecasting and Generation with Conditional Flow Matching

Sean Ye, M. Gombolay

GenerationAutonomous DrivingDiffusion modelFlow-based ModelTime SeriesSequential

🎯 What it does: Proposed Trajectory Conditional Flow Matching (T-CFM), unifying trajectory prediction and generation tasks, utilizing flow matching techniques to learn time-varying vector fields, achieving efficient and fast trajectory generation and prediction.

Efficient-PIP: Large-scale Pixel-level Aligned Image Pair Generation for Cross-time Infrared-RGB Translation

Jian Li, Zhenping Sun

Image TranslationImagePoint Cloud

🎯 What it does: Proposed the PIP framework for efficiently generating pixel-level aligned data pairs of cross-temporal infrared-RGB images, and constructed the NUDT-PIP dataset;

Efficiently Obtaining Reachset Conformance for the Formal Analysis of Robotic Contact Tasks

Chencheng Tang, Matthias Althoff

OptimizationRobotic Intelligence

🎯 What it does: Generate a hybrid automaton model that conforms to reachable sets to support formal verification of robot contact tasks, capturing recorded behaviors by injecting non-determinism into continuous and discrete dynamics, and optimizing the identification of all model parameters and non-determinism.

EgoVM: Achieving Precise Ego-Localization using Lightweight Vectorized Maps

Yuzhe He, Guowei Wan

Autonomous DrivingTransformerSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Proposed EgoVM, an end-to-end network that utilizes a lightweight vectorized map to achieve precise self-vehicle localization.

Elliptical K-Nearest Neighbors - Path Optimization via Coulomb’s Law and Invalid Vertices in C-space Obstacles

Liding Zhang, A. Knoll

OptimizationRobotic IntelligencePhysics Related

🎯 What it does: Proposed a sampling-based path planner FDIT*, which utilizes invalid vertex information and Coulomb's law, integrating elliptical k-nearest neighbor search to achieve rapid convergence.

Embedded 3D Printing of Silicone for Soft Actuator with Stiffness Gradient and Programmable Workspace

Fei Xiao, Jian Zhu

Robotic Intelligence

🎯 What it does: Using an embedded 3D printing method with a single active hybrid print head, the hardness gradient of two-component silicone rubber was controlled by adjusting mixing ratios, successfully fabricating a soft pneumatic actuator with a cantilever structure, achieving a programmable workspace for elongation, radial expansion, and bending.

Embedded Valves for Distributed Control of Soft Pneumatic Actuators

Runze Zuo, Daniel Bruder

Robotic Intelligence

🎯 What it does: Developed an embedded miniature pressure regulation system to achieve distributed pressure control for multiple actuators inside soft pneumatic McKibben actuators.

Embodied AI with Two Arms: Zero-shot Learning, Safety and Modularity

Jacob Varley, Vikas Sindhwani

Robotic IntelligenceTransformerLarge Language ModelTextPoint Cloud

🎯 What it does: Proposed an embodied AI system that can receive open-ended natural language instructions from humans and accomplish tasks with a long time span and a large workspace through collaborative dual-arm operations.

Embodied Uncertainty-Aware Object Segmentation

Xiaolin Fang, Tomás Lozano-Pérez

SegmentationRobotic IntelligenceLarge Language ModelImage

🎯 What it does: Proposes an uncertainty-aware target instance segmentation method (UncOS) and demonstrates its effectiveness in embodied interactive segmentation.

Embodiment Randomization for Cross Embodiment Navigation

Pranav Putta, Arjun Majumdar

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes a technique called Embodiment Randomization for training robust navigation strategies that can transfer across different robot body forms.

Embodiment: Self-Supervised Depth Estimation Based on Camera Models

Jinchang Zhang, Guoyu Lu

Depth Estimation

🎯 What it does: Embed the physical attributes of a camera's intrinsic and extrinsic parameters into a monocular depth estimation model, utilizing these attributes to calculate depth priors for the ground and related areas, thereby providing unsupervised supervision signals without the need for additional sensors.

EMBOSR: Embodied Spatial Reasoning for Enhanced Situated Question Answering in 3D Scenes

Yu Hao, Yu-Shen Liu

GenerationTransformerLarge Language ModelVision Language ModelVision-Language-Action ModelChain-of-Thought

🎯 What it does: Proposes a novel spatial reasoning paradigm in 3D scenes that combines base models with chain-of-thought reasoning, aiming to enhance performance on the SQA3D task and extend to scenarios such as 3D caption generation

Emotional Tandem Robots: How Different Robot Behaviors Affect Human Perception While Controlling a Mobile Robot

Julian Kaduk, Heiko Hamann

Robotic Intelligence

🎯 What it does: The study investigates the impact of emotional expression actions by follower robots on human operators' perception in a leader-follower robot scenario.

Empathetic Response Generation System: Enhancing Photo Reminiscence Chatbot with Emotional Context Analysis

Alberto Herrera Ruiz, Li-Chen Fu

TransformerTextRetrieval-Augmented Generation

🎯 What it does: Built an empathy response generation system to enhance photo memory chatbots that only ask questions, improving dialogue engagement through emotional context analysis.

EMPOWER: Embodied Multi-role Open-vocabulary Planning with Online Grounding and Execution

F. Argenziano, D. Bloisi

Robotic IntelligenceLarge Language Model

🎯 What it does: Proposes the EMPOWER framework to achieve open-vocabulary online grounding and planning for robots in real-world environments, implemented through a multi-role mechanism.

Enabling Maintainablity of Robot Programs in Assembly by Extracting Compositions of Force- and Position-Based Robot Skills from Learning-from-Demonstration Models

Daniel Andreas Bargmann, Marco F. Huber

Robotic Intelligence

🎯 What it does: Create maintainable hierarchical skill sequences by extracting robot skills combining force and position control from a learning demonstration model, allowing experts to manually adjust without requiring re-demonstration.

EnduRL: Enhancing Safety, Stability, and Efficiency of Mixed Traffic Under Real-World Perturbations Via Reinforcement Learning

Bibek Poudel, K. Heaslip

Autonomous DrivingOptimizationReinforcement LearningTime Series

🎯 What it does: Acceleration curves are extracted by analyzing real driving trajectories, and a reinforcement learning-based robot vehicle controller is trained and evaluated in a mixed traffic simulation with ring and bottleneck scenarios, focusing on safety, efficiency, and stability.

Energy Minimization using Custom-Designed Magnetic-Spring Actuators

Y. Fu, David J. Braun

OptimizationRobotic IntelligencePhysics Related

🎯 What it does: This study proposes a novel actuator by combining a standard motor (uniform magnetic field) with a custom rotating magnetic spring (non-uniform magnetic field), developing a prototype and providing a systematized computational method to customize the magnetic field to minimize energy consumption in user-defined oscillatory tasks.

Energy Sharing Mechanism for Freeform Robots Utilizing Conductive Spherical Sliding Surfaces

Xinzhuo Li, T. Lam

Robotic Intelligence

🎯 What it does: Proposes an energy-sharing mechanism for the FreeSN modular self-reconfigurable robot on conductive spherical sliding surfaces, utilizing face-to-face connections through brush contact and shell decomposition to establish energy-sharing channels at any point on the spherical surface.

Energy-efficient Trajectory Planning with Media Transition for a Hybrid Unmanned Aerial-Underwater Vehicle

P. Pinheiro, Paulo L. J. Drews

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Proposed a hybrid trajectory planning method for unmanned aerial vehicles (UAVs) and underwater vehicles based on RRT, and designed two new heuristic strategies

Energy-Optimal Planning of Waypoint-Based UAV Missions - Does Minimum Distance Mean Minimum Energy?

Nicolas Michel, Xinfan Lin

Optimization

🎯 What it does: This study investigates waypoint sequence planning for multi-rotor drones in 3D space based on energy consumption minimization, proposing an energy-optimal planning method and comparing it with the traditional shortest distance Traveling Salesman Problem (TSP).

Energy-Optimized Planning in Non-Uniform Wind Fields with Fixed-Wing Aerial Vehicles

Yufei Duan, Roland Siegwart

OptimizationPhysics Related

🎯 What it does: Propose a sampling-based planner that utilizes ground-relative Dubins kinematic paths to plan energy-optimal paths in non-uniform wind fields.

Enhanced Language-guided Robot Navigation with Panoramic Semantic Depth Perception and Cross-modal Fusion

Liuyi Wang, Qi Chen

Robotic IntelligenceTransformerVision Language ModelVision-Language-Action ModelMultimodality

🎯 What it does: Proposed the SEAT model, utilizing panoramic multi-type visual encoder, region query pre-training task, and dual-scale cross-modal Transformer to achieve semantic-depth aware cross-modal navigation;

Enhanced Model Robustness to Input Corruptions by Per-corruption Adaptation of Normalization Statistics

Elena Camuffo, Mete Ozay

Adversarial AttackConvolutional Neural NetworkImageBenchmark

🎯 What it does: Proposes the Per-corruption Adaptation of Normalization statistics (PAN) method to enhance the robustness of visual systems under input corruptions.

Enhanced Omni-Ball: Spherical Omnidirectional Wheel Achieving Passive Rollers with High Load Capacity and Smoothness through an Offset Rotational Axis

K. Tadakuma, Satoshi Tadokoro

🎯 What it does: Designed and manufactured a spherical omnidirectional wheel with an offset rotating axis to achieve high load capacity and smoothness for passive wheels;

Enhanced Robotic Assistance for Human Activities through Human-Object Interaction Segment Prediction

Yuankai Wu, Eckehard G. Steinbach

Robotic IntelligenceGraph Neural NetworkVision-Language-Action ModelSimultaneous Localization and MappingImageVideo

🎯 What it does: Propose a two-stage robot-assisted system that infers human intent using future human-object interaction (HOI) segment prediction and completes tasks through visual navigation, target localization, and grasping.

Enhancing 3D Single Object Tracking with Efficient Point Cloud Segmentation

Yushi Yang, Hongxin Xu

Object TrackingAutonomous DrivingConvolutional Neural NetworkPoint CloudBenchmark

🎯 What it does: Proposed an efficient 3D single-object tracker (EST) that can effectively segment point cloud features.

Enhancing Exploratory Capability of Visual Navigation Using Uncertainty of Implicit Scene Representation

Yichen Wang, Hesheng Wang

Autonomous DrivingNeural Radiance FieldImage

🎯 What it does: Proposed an uncertainty-driven exploration navigation pipeline (NUE) based on NeRF, leveraging NeRF's uncertainty to enhance exploration behavior and integrating memory information to generate navigation actions.

Enhancing Leg Odometry in Legged Robots with Learned Contact Bias: An LSTM Recurrent Neural Network Approach

Yaru Gu, Ting Zou

Robotic IntelligenceRecurrent Neural NetworkSimultaneous Localization and MappingPoint CloudTime Series

🎯 What it does: Learn the robot's foot contact deviation using LSTM RNN and fuse it into a Kalman filter to improve leg odometry accuracy in real-time.

Enhancing LiDAR Scene Upsampling with Instance-aware Feature-embedding and Attention Mechanism

Weijie Wang, Chieh-Chih Wang

Super ResolutionTransformerPoint Cloud

🎯 What it does: Propose a LiDAR point cloud super-sampling method based on instance embedding auxiliary tasks and context attention modules, aiming to reduce artifacts in multi-target scenarios and improve sampling quality.

Enhancing Nighttime UAV Tracking with Light Distribution Suppression

Liangliang Yao, Kunhan Lu

RestorationObject TrackingVideo

🎯 What it does: Proposed LDEnhancer, a light distribution suppression enhancer for nighttime UAV tracking, addressing the problem of over-enhancement caused by uneven illumination in low-light images.

Enhancing Object Grasping Efficiency with Deep Learning and Post-processing for Multi-finger Robotic Hands

Pouya Samandi, M. Mehrandezh

Robotic IntelligenceImage

🎯 What it does: Improved the Grasp-Rectangle method to support grasping from top and side views on multi-fingered grippers, and computed optimal grasp poses and multi-point contacts through geometric cues.

Enhancing Online Road Network Perception and Reasoning with Standard Definition Maps

Hengyuan Zhang, Liu Ren

Autonomous DrivingImageGraph

🎯 What it does: Leverage standard definition (SD) maps as a prior, combining rasterized SD map representations with various online mapping architectures, and extend the OpenLane-V2 dataset to OpenStreetMaps to evaluate the advantages of graphical SD map representations.

Enhancing Prosthetic Safety and Environmental Adaptability: A Visual-Inertial Prosthesis Motion Estimation Approach on Uneven Terrains

Chuheng Chen, Chenglong Fu

Robotic IntelligenceSimultaneous Localization and MappingOptical FlowImage

🎯 What it does: Propose a visual inertial motion estimation method combining depth camera and error-state Kalman filter for motion perception and control of prosthetics on uneven ground

Enhancing Reinforcement Learning in Sensor Fusion: A Comparative Analysis of Cubature and Sampling-based Integration Methods for Rover Search Planning

Jan-Hendrik Ewers, Douglas G. Thomson

Autonomous DrivingOptimizationReinforcement Learning

🎯 What it does: Compared the computational speed and accuracy of the cube integration method and the sampling integration method in 2D polygon integration.

Enhancing Robustness in Manipulability Assessment: The Pseudo-Ellipsoid Approach

Erfan Shahriari, Sami Haddadin

Robotic Intelligence

🎯 What it does: Proposed and implemented the manipulability pseudo-ellipsoid method to evaluate the kinematic or mechanical manipulability of a joint system in specific configurations.

Enhancing Safety via Deep Reinforcement Learning in Trajectory Planning for Agile Flights in Unknown Environments

Lidia Rocha, K. Vivaldini

Robotic IntelligenceSupervised Fine-TuningReinforcement Learning

🎯 What it does: Combining supervised learning teacher strategies with deep reinforcement learning student strategies enhances the ability of UAVs to perform rapid and safe trajectory planning in unknown environments.

Enhancing Surgical Precision in Autonomous Robotic Incisions via Physics-Based Tissue Cutting Simulation

J. Ge, Axel Krieger

Robotic IntelligenceMeshPhysics Related

🎯 What it does: Utilize finite element-based tissue cutting simulation with autonomous robotic surgery to pre-predict and compensate for tissue deformation during cutting, thereby improving incision accuracy.

Enhancing Visual Place Recognition via Fast and Slow Adaptive Biasing in Event Cameras

Gokul B. Nair, Tobias Fischer

Recognition

🎯 What it does: This paper proposes an adaptive bias regulation feedback control algorithm for event cameras to enhance visual localization performance.

Ensembling Prioritized Hybrid Policies for Multi-agent Pathfinding

Huijie Tang, Jinkyoo Park

OptimizationReinforcement LearningMixture of Experts

🎯 What it does: Proposed and implemented the Ensembling Prioritized Hybrid Policies (EPH) method, which enhances performance in multi-agent path planning through a selective communication module, Q-learning training, integration of neural policies with single-intelligent expert guidance, Q-value prioritized conflict resolution, and robust ensemble methods.

Ensuring Joint Constraints of Torque-Controlled Robot Manipulators under Bounded Jerk

Dongwoo Ko, Wan Kyun Chung

OptimizationRobotic Intelligence

🎯 What it does: Proposes an optimization-based control framework that can satisfy robot joint position, velocity, and acceleration constraints under limited angular jerk;

Ensuring Safety in LLM-Driven Robotics: A Cross-Layer Sequence Supervision Mechanism

Ziming Wang, Man Li

Safty and PrivacyRobotic IntelligenceLarge Language Model

🎯 What it does: Propose a cross-layer sequential supervision mechanism, utilizing linear temporal logic and Büchi automata to real-time monitor and correct task planning and action execution driven by LLMs, enhancing safety.

Environment Transformer and Policy Optimization for Model-Based Offline Reinforcement Learning

Pengqin Wang, S. Shen

OptimizationTransformerReinforcement LearningWorld Model

🎯 What it does: Proposes an Environment Transformer based on the Transformer architecture for dynamic and reward function modeling in model-free offline reinforcement learning, combined with Conservative Q-Learning (CQL).

Environment-Adaptive Gait Planning for Obstacle Avoidance in Lower-Limb Robotic Exoskeletons

Edoardo Trombin, L. Tonin

Pose EstimationRobotic Intelligence

🎯 What it does: Proposed an environmentally adaptive gait planning method for lower-limb exoskeleton robots in obstacle environments.

Environmental and Behavioral Imitation for Autonomous Navigation

J. Aoki, Ryo Kurazume

Autonomous DrivingRobotic IntelligenceNeural Radiance FieldImage

🎯 What it does: Proposes a framework that utilizes NeRF to generate simulated environments for navigation imitation learning, enabling strategy learning from single images without requiring a physical robot to complete the expert teaching and learning process, and transferring the learned strategies to real robots.

Equivariant Ensembles and Regularization for Reinforcement Learning in Map-based Path Planning

Mirco Theile, Alberto L. Sangiovanni-Vincentelli

Autonomous DrivingReinforcement Learning

🎯 What it does: Propose an equivariant ensemble method that does not require specialized network components, constructing equivariant policies and equivariant-invariant value functions, and adding a regularization term to enhance inductive bias during training, significantly improving sample efficiency and performance in map-based path planning cases.

Error-State Kalman Filter based Visual-Inertial Odometry Using Orientation Measurement on Unit Quaternion Group

Chao Chang, Feng-Li Lian

Pose EstimationSimultaneous Localization and MappingMultimodality

🎯 What it does: Derive an innovation signal on the unit quaternion group (S³) for attitude measurement within the error-state Kalman filter (ESKF) framework.

ESO-SLAM: Tightly-Coupled and Simultaneous Estimation of Self and Multi-Object Pose via Sensor Fusion

Wu Li, Guiyuan Wang

Pose EstimationAutonomous DrivingOptimizationRobotic IntelligenceSimultaneous Localization and MappingOptical FlowVideoPoint Cloud

🎯 What it does: Proposed the ESO-SLAM system, achieving tightly coupled simultaneous pose estimation for self-robots and multiple dynamic targets through multi-sensor fusion, including a multi-probability fusion tracker, a dynamic point cloud removal method decoupling 3D Kalman filter velocity prior with camera optical flow, and a framework for joint optimization of pose using multi-constraint factors.

Estimating Perceptual Uncertainty to Predict Robust Motion Plans

Arjun Gupta, Saurabh Gupta

Autonomous DrivingRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Developed a framework for predicting perceptual uncertainty in neural network-based vision models, and applied it to robust motion planning in mobile manipulation tasks.

Estimating the Joint Angles of a Magnetic Surgical Tool using Monocular 3D Keypoint Detection and Particle Filtering

Erik Fredin, Eric D. Diller

Pose EstimationConvolutional Neural NetworkImageBiomedical Data

🎯 What it does: Using a monocular endoscopic camera, combined with a deep 3D keypoint detection framework and a particle filter, to real-time estimate the pose and joint angles of magnetic joint surgical instruments.

Evaluating a Movable Palm in Caging Inspired Grasping using a Reinforcement Learning-based Approach

Luke Beddow, Dimitrios Kanoulas

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: The study investigates the effectiveness of using a rigid movable palm for grasping multiple objects, and compares the impact of having or not having a palm, different finger stiffness, and fingertip angles on grasping success rate and stability.

Evaluating Dynamic Environment Difficulty for Obstacle Avoidance Benchmarking

Moji Shi, Javier Alonso-Mora

Autonomous DrivingBenchmark

🎯 What it does: Propose and evaluate four metrics for quantifying dynamic environment difficulty, and validate their effectiveness through large-scale simulation experiments.

Evaluating Gait Symmetry with a Smart Robotic Walker: A Novel Approach to Mobility Assessment

Mahdi Chalaki, M. Tavakoli

Robotic IntelligenceTime Series

🎯 What it does: Developed and validated a gait symmetry detection method based on an intelligent robotic walker, which extracts gait features and calculates symmetry indices by analyzing interaction torque signals between the user and the walker.

Evaluating the Impact of a Semi-Autonomous Interface on Configuration Space Accessibility for Multi-DOF Upper Limb Prostheses

Rebecca J. Greene, Nitish V. Thakor

Robotic IntelligenceMultimodalityBiomedical Data

🎯 What it does: Evaluate the impact of a semi-automated hybrid gaze-EMG interface on the configuration space accessibility of multi-degree-of-freedom upper-limb prostheses, and compare its performance with traditional EMG classification controllers across different mechanical degrees of freedom.

Evaluation and Deployment of LiDAR-based Place Recognition in Dense Forests

Haedam Oh, Maurice F. Fallon

RecognitionPose EstimationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Analyzes the performance of four LiDAR location recognition systems in dense forests, and integrates the best Logg3dNet into a complete 6-DoF pose estimation system, demonstrating the effectiveness of three operational modes: online SLAM, offline multi-task SLAM map merging, and relocalization.

Evaluation and Design Recommendations for a Folding Morphing-wheg Robot for Nuclear Characterisation

Dominic Murphy, Paul A. Bremner

Robotic IntelligencePhysics Related

🎯 What it does: Designed, manufactured, and tested a foldable robot with morphing-whegs that can pass through a 150 mm diameter entrance pipe for mapping and characterization inside nuclear facilities.

Evaluation of Predictive Display for Teleoperated Driving Using CARLA Simulator

Fatima Kashwani, Jorge Dias

Object DetectionSegmentationAutonomous DrivingTransformer

🎯 What it does: Evaluates the performance of a prediction display method based on free space-guided paths in vehicle remote control.

Evaluation of the Design of a Tool for the Automated Assembly of Preconfigured Wires*

Stefanie Bartelt, B. Kuhlenkötter

OptimizationRobotic Intelligence

🎯 What it does: Discussed and designed a scheme for an automated pre-made cable assembly tool, and verified it on a demonstration machine.

Event-based Few-shot Fine-grained Human Action Recognition

Zonglin Yang, Liyuan Pan

RecognitionMeta LearningImageMultimodalityTime Series

🎯 What it does: Proposes a fine-grained human action recognition approach based on event cameras, constructs the first event camera fine-grained human action dataset E-FAction, and designs an event feature extractor guided by RGB frames to achieve robust weight initialization in few-shot fine-grained action recognition.

Event-Free Moving Object Segmentation from Moving Ego Vehicle

Zhuyun Zhou, D. Ginhac

SegmentationAutonomous DrivingTransformerBenchmark

🎯 What it does: Propose using an event camera for moving object segmentation, construct the first large-scale dataset DSEC-MOS, and design the EmoFormer network.

Event-intensity Stereo with Cross-modal Fusion and Contrast

Yuanbo Wang, Xin Yang

Depth EstimationRepresentation LearningContrastive LearningMultimodality

🎯 What it does: By introducing a cross-modal fusion module, learning visual representations and combining event-intensity features through contrastive learning to enhance the accuracy of stereo matching and disparity estimation

Every Dataset Counts: Scaling up Monocular 3D Object Detection with Joint Datasets Training

Fulong Ma, Ming Liu

Object DetectionAutonomous DrivingImagePoint Cloud

🎯 What it does: Train a monocular 3D object detection model combining 3D and 2D datasets, and propose a robust model adaptable to different camera setups, selective training strategies, and pseudo-3D training methods.

EverySync: An Open Hardware Time Synchronization Sensor Suite for Common Sensors in SLAM

Xuankang Wu, Zheng Fang

Simultaneous Localization and MappingImageMultimodalityPoint CloudTime Series

🎯 What it does: Proposed EverySync, an open-source hardware time synchronization system for multi-sensor (camera, IMU, LiDAR, GNSS/RTK) synchronization.

Evidential Semantic Mapping in Off-road Environments with Uncertainty-aware Bayesian Kernel Inference

Junyoung Kim, Jihong Min

SegmentationAutonomous DrivingConvolutional Neural Network

🎯 What it does: Proposes a semantic mapping method based on evidential deep learning, integrating semantic uncertainty estimation into uncertainty-aware Bayesian Kernel Inference (BKI) to construct more reliable semantic maps in offline environments.

EVIT: Event-based Visual-Inertial Tracking in Semi-Dense Maps Using Windowed Nonlinear Optimization

Runze Yuan, L. Kneip

Pose EstimationOptimizationSimultaneous Localization and MappingMultimodality

🎯 What it does: Propose incorporating inertial signals into event camera semi-dense map tracking, achieving more robust pose estimation through windowed nonlinear optimization.

EVSMap: An Efficient Volumetric-Semantic Mapping Approach for Embedded Systems

Jiyuan Qiu, Haowen Wang

SegmentationComputational EfficiencyConvolutional Neural NetworkSimultaneous Localization and MappingImage

🎯 What it does: Proposed EVSMap, an efficient real-time volumetric-semantic mapping framework, which includes a lightweight RGB-D semantic segmentation network and three improved modules, and introduces an enhanced multi-class Bayesian update strategy to reduce memory usage and improve mapping speed.

Expansion-GRR: Efficient Generation of Smooth Global Redundancy Resolution Roadmaps

Zhuoyun Zhong, Constantinos Chamzas

OptimizationRobotic Intelligence

🎯 What it does: Developed an algorithm called EXPANSION-GRR that rapidly generates smooth global redundant resolution roadmap (GRR) by utilizing efficient configuration space projection combined with multiple seeding strategies.

Experience-Learning Inspired Two-Step Reward Method for Efficient Legged Locomotion Learning Towards Natural and Robust Gaits

Yinghui Li, Yufei Xue

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose a two-step reward framework, first learning basic gait on flat ground, then leveraging the learned experience for adversarial imitation learning on complex terrains to achieve natural and stable gait.

Explainable Artificial intelligence for Autonomous UAV Navigation

Didula Dissanayaka, R. Gosine

Explainability and InterpretabilityRobotic IntelligenceReinforcement LearningImageText

🎯 What it does: Propose a UAV autonomous navigation method based on explainable AI, combining SAC+MLP deep reinforcement learning controller with a moving window gradient XAI framework, providing visual and textual explanations;

Explicit Interaction for Fusion-Based Place Recognition

Jingyi Xu, Ling Pei

RecognitionAutonomous DrivingSupervised Fine-TuningImageMultimodalityPoint CloudBenchmark

🎯 What it does: Proposed the EINet network, achieving explicit interaction between two modalities for location recognition based on fusion.

Exploiting Hybrid Policy in Reinforcement Learning for Interpretable Temporal Logic Manipulation

Hao Zhang, Zhen Kan

Explainability and InterpretabilityReinforcement Learning

🎯 What it does: Proposes a hybrid strategy framework called HyTL based on linear temporal logic (LTL), utilizing a three-layer decision structure to enhance the performance and explainability of reinforcement learning (RL) in complex operational tasks.

Exploiting Local Features and Range Images for Small Data Real-Time Point Cloud Semantic Segmentation

Daniel Fusaro, Alberto Pretto

SegmentationComputational EfficiencyPoint Cloud

🎯 What it does: Proposes a point cloud semantic segmentation method that combines local features from 3D representations with range image-based representations, utilizing GPU-accelerated KDTree for fast construction, querying, and projection.

Exploiting Priors from 3D Diffusion Models for RGB-Based One-Shot View Planning

Sicong Pan, Maren Bennewitz

OptimizationDiffusion modelImage

🎯 What it does: Propose an RGB one-shot viewpoint planning method based on a single RGB image, leveraging the geometric prior of a 3D diffusion model to achieve effective viewpoint planning and global shortest path search.

Exploratory Motion Guided Tactile Learning for Shape-Consistent Robotic Insertion

Gang Yan, Shigeki Sugano

Robotic IntelligenceTransformerTime SeriesSequential

🎯 What it does: Propose exploration actions and a transformer-based neural network to estimate and compensate for residual position uncertainty in robotic insertion tasks through tactile perception.

Exploring 3D Human Pose Estimation and Forecasting from the Robot’s Perspective: The HARPER Dataset

Andrea Avogaro, Marco Cristani

Pose EstimationVideoBenchmark

🎯 What it does: Proposed and constructed the HARPER dataset, focusing on 3D human pose estimation and prediction from the robot Spot's perspective;

Exploring Cognitive Load Dynamics in Human-Machine Interaction for Teleoperation: A User-Centric Perspective on Remote Operation System Design

Juan Jose Garcia Cardenas, Adriana Tapus

Robotic IntelligenceTime SeriesBiomedical Data

🎯 What it does: This study conducted teleoperation experiments with 20 participants under three conditions (on-site, remote visual feedback, and remote panoramic robot) to explore the impact of information availability on operators' cognitive load, physiological responses (skin conductance response, blink rate, facial temperature), and task performance.

Exploring Constrained Reinforcement Learning Algorithms for Quadrupedal Locomotion

Joonho Lee, Marco Hutter

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Evaluated five constrained policy optimization algorithms for motion control on three different quadruped robot models using the Constrained Markov Decision Process (CMDP) framework, with a focus on sim-to-real transfer under real-world physical constraints.

Exploring Few-Beam LiDAR Assistance in Self-Supervised Multi-Frame Depth Estimation

Rizhao Fan, Stefano Mattoccia

Depth EstimationPoint Cloud

🎯 What it does: Propose a method that integrates sparse depth information from few-beam LiDAR with multi-frame matching to enhance the robustness of self-supervised multi-frame depth estimation.

Exploring How Non-Prehensile Manipulation Expands Capability in Robots Experiencing Multi-Joint Failure

Gilberto Briscoe-Martinez, Alessandro Roncone

Robotic Intelligence

🎯 What it does: The study explores using non-grasping manipulation and whole-body interaction to complete manipulation tasks on robots with multiple joint failures.

Exploring Latent Pathways: Enhancing the Interpretability of Autonomous Driving with a Variational Autoencoder

Anass Bairouk, Daniela Rus

Autonomous DrivingExplainability and InterpretabilityAuto EncoderImage

🎯 What it does: Combine variational autoencoder with bio-inspired neural circuit policy controller to directly generate steering commands from camera images, and propose an automatic latent variable perturbation tool for interpreting model internal decisions

Exploring Modal Switch in Metamaterial-Based Robots

Britton Jordan, Alan Kuntz

Robotic Intelligence

🎯 What it does: Developed a multi-modal bending/shear and bending/torsion mechanical deformation material robot prototype capable of switching between two different forms, with a detailed description of its design and the sliding rod mechanical switching mechanism.

Explosive Legged Robotic Hopping: Energy Accumulation and Power Amplification via Pneumatic Augmentation

Yifei Chen, Xiaobin Xiong

OptimizationRobotic Intelligence

🎯 What it does: Designed and implemented a pneumatic boosting system on a traditional motor-driven legged robot to achieve intermittent high-power burst jumps, and validated its effectiveness on a homemade single-leg jumping robot.

Exposing the Unseen: Exposure Time Emulation for Offline Benchmarking of Vision Algorithms

Olivier Gamache, Philippe Giguère

Auto EncoderSimultaneous Localization and MappingImagePoint CloudBenchmark

🎯 What it does: Propose an exposure time simulator that generates images with arbitrary exposure times to enable offline benchmark testing of visual algorithms.

Extended Tree Search for Robot Task and Motion Planning

Tianyu Ren, Jan Peters

OptimizationRobotic Intelligence

🎯 What it does: Proposes a TAMP decision framework based on an extended decision tree, integrating symbolic task planning and high-dimensional motion variable binding, ultimately solving for the optimal solution through MCTS.

Extending Task and Motion Planning with Feasibility Prediction: Towards Multi-Robot Manipulation Planning of Realistic Objects

Smail Ait Bouhsain, Thierry Siméon

Robotic Intelligence

🎯 What it does: Proposed a multi-robot task and motion planning algorithm based on feasibility prediction, and extended it to grid-shaped objects

Extensive, Long-term Task and Motion Planning with Signal Temporal Logic Specification for Autonomous Construction

Mineto Satoh, Hiroyuki Oyama

OptimizationRobotic Intelligence

🎯 What it does: Proposes a hierarchical task and motion planning (TAMP) method for deformable objects (e.g., earth excavation) in autonomous construction, generating efficient task plans that satisfy high-level construction goals and are motion-feasible.

External Interaction Estimation of 6-PSS Parallel Robots with Embodied Mechanical Intelligence

Jingyuan Xia, A. Gao

OptimizationRobotic IntelligenceTime Series

🎯 What it does: A method for estimating external interaction of a 6-PSS parallel robot using embedded mechanical intelligence is proposed. Dual single-axis force sensors are embedded in each leg, combined with drive motor encoder information, and a backpropagation neural network is used to simultaneously estimate external force and its position.

Extrinsic Calibration of Multiple LiDARs for a Mobile Robot based on Floor Plane And Object Segmentation

Shun Niijima, Masaya Kinoshita

SegmentationPose EstimationAutonomous DrivingPoint Cloud

🎯 What it does: Proposes a target-free extrinsic calibration method based on ground plane and object segmentation for mobile robots equipped with multiple LiDARs with non-overlapping fields of view.

EyeSight Hand: Design of a Fully-Actuated Dexterous Robot Hand with Integrated Vision-Based Tactile Sensors and Compliant Actuation

Branden Romero, Edward H. Adelson

Robotic IntelligenceImage

🎯 What it does: Designed a 7-degree-of-freedom humanoid hand called EyeSight Hand, and evaluated it on three challenging tasks: bottle opening, plasticine cutting, and plate picking and placing; simultaneously, a vision-based tactile sensor and a quasi-direct drive compliant actuation scheme were used to train an imitative learning model with a visual loss strategy;

Fast and Communication-Efficient Multi-UAV Exploration Via Voronoi Partition on Dynamic Topological Graph

Qianli Dong, Xuebo Zhang

OptimizationRobotic IntelligenceGraph

🎯 What it does: Proposed a fast and communication-efficient multi-robot exploration method using Multi-Robot Dynamic Topology Graph (MR-DTG) and achieving task allocation based on graph Voronoi partitioning

Fast Explicit-Input Assistance for Teleoperation in Clutter

Nick Walker, Claudia P'erez-D'Arpino

OptimizationRobotic Intelligence

🎯 What it does: Propose an explicit pointing-based teleoperation assistance interface for performing multi-step separation and stacking tasks in crowded environments.

Fast Global Point Cloud Registration using Semantic NDT

Robert Schirmer, C. Stachniss

Pose EstimationRepresentation LearningPoint Cloud

🎯 What it does: A fast global point cloud registration method is proposed, which estimates the rigid transformation between the source point cloud and the target point cloud using semantic NDT.

Fast Spatial Reasoning of Implicit 3D Maps through Explicit Near-Far Sampling Range Prediction

Chaerin Min, Jongwoo Lim

Depth EstimationComputational EfficiencyRobotic IntelligencePoint Cloud

🎯 What it does: Propose an adaptive prediction method for depth mapping near-far sampling range, and design an algorithm for automatically detecting failed predictions and recovery.

FBG-based Shape-Sensing to Enable Lateral Deflection Methods of Autonomous Needle Insertion

Dimitri A. Lezcano, Jin Seob Kim

Robotic IntelligenceBiomedical Data

🎯 What it does: Extended the Lie-group theory-based shape perception model to handle lateral deflection of injection needles during insertion, and validated it through robotic insertion experiments in simulated tissue.