arXivSub Start free trial

ICRA 2024 Papers — Page 12

IEEE International Conference on Robotics and Automation · 1760 papers

Observer-based Distributed MPC for Collaborative Quadrotor-Quadruped Manipulation of a Cable-Towed Load

Shaohang Xu, Lijun Zhu

OptimizationRobotic Intelligence

🎯 What it does: Propose a collaborative quadrotor-quadruped robot system for manipulating tethered loads, addressing the unknown dynamics of the load by building a new nonlinear dynamic model, designing an observer, and integrating it into a distributed MPC framework.

OCC-VO: Dense Mapping via 3D Occupancy-Based Visual Odometry for Autonomous Driving

Heng Li, Yanyong Zhang

Autonomous DrivingTransformerSimultaneous Localization and MappingImage

🎯 What it does: Propose the OCC-VO framework, which uses deep learning to convert 2D camera images into 3D semantic occupancy grids, achieving visual odometry and global semantic map construction without requiring traditional pose and feature point estimation.

Occluded Part-aware Graph Convolutional Networks for Skeleton-based Action Recognition

Min Hyuk Kim, S. Yoo

RecognitionGraph Neural NetworkAuto EncoderGraph

🎯 What it does: Proposes an Occlusion-Aware Graph Convolutional Network (OP-GCN) for action recognition when human skeletons are occluded

ODD-based Query-time Scenario Mutation Framework for Autonomous Driving Scenario databases

Yun Tang, P. Jennings

Autonomous Driving

🎯 What it does: Propose a query-time scenario mutation framework based on ODD tags to improve query efficiency and diversity in large autonomous driving scenario databases.

Odometry Estimation by Fusing Multiple Radar Sensors and an Inertial Measurement Unit

Tim Brühl, Sören Hohmann

Autonomous DrivingSimultaneous Localization and MappingMultimodalityTime Series

🎯 What it does: Proposes an automotive odometry estimation framework using six asynchronous millimeter-wave radar sensors along with a gyroscope and accelerometer.

Offline Goal-Conditioned Reinforcement Learning for Safety-Critical Tasks with Recovery Policy

Chenyang Cao, Xueqian Wang

Reinforcement Learning

🎯 What it does: Proposes a method called Recovery-based Supervised Learning (RbSL) to address offline goal-conditional reinforcement learning tasks with safety constraints, aiming to achieve multi-objective safety-critical missions.

OmniColor: A Global Camera Pose Optimization Approach of LiDAR-360Camera Fusion for Colorizing Point Clouds

Bonan Liu, Pan Hui

Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingImageMultimodalityPoint Cloud

🎯 What it does: Propose OmniColor, which utilizes an independent 360-degree camera to jointly optimize camera poses through photometric consistency optimization, mapping images to LiDAR-based point clouds to achieve point cloud coloring.

Omnidirectional Dense SLAM for Back-to-back Fisheye Cameras

Weijian Xie, Guofeng Zhangv

Depth EstimationOptimizationSimultaneous Localization and MappingOptical FlowImageMultimodality

🎯 What it does: Proposes a real-time visual-inertial dense SLAM system, utilizing back-to-back dual fisheye cameras to achieve 360° environmental coverage, incorporating a sliding window frontend and a lightweight panoramic depth completion network based on multi-base depth representation.

OmniLRS: A Photorealistic Simulator for Lunar Robotics

Antoine Richard, Kazuya Yoshida

SegmentationData SynthesisDomain AdaptationData-Centric LearningRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Propose OmniLRS—a lighting-realistic lunar simulator based on the Nvidia robotics simulator, offering fast procedural environment generation, multi-robot capabilities, synthetic data pipelines for machine learning, and support for ROS1/ROS2 bindings.

On camera model conversions

Eva Goichon, Fumio Kanehiro

Simultaneous Localization and MappingBenchmark

🎯 What it does: Proposes conversion methods between three camera models, enabling the mutual conversion of data calibrated under different projection models, allowing the use of existing data without recalibration;

On Experimental Emulation of Printability and Fleet Aware Generic Mesh Decomposition for Enabling Aerial 3D Printing

Marios-Nektarios Stamatopoulos, G. Nikolakopoulos

OptimizationRobotic IntelligenceMeshPhysics Related

🎯 What it does: The feasibility of a block-based multi-degree-of-freedom aerial drone 3D printing framework was verified through experimental simulation, demonstrating precise motion planning and task allocation.

On Robust Control Laws Trade-off Analysis for Space Manipulators with Uncertain Parameters and Flexible Appendages*

Kostas Nanos, E. Papadopoulos

Robotic IntelligencePhysics Related

🎯 What it does: Studied and compared robust controllers for space manipulators under parameter uncertainty, noisy measurements, and disturbances, with a focus on evaluating the performance of LPV+H∞ controllers against model-based PD controllers and PD+H∞ combinations.

On the Disentanglement of Tube Inequalities in Concentric Tube Continuum Robots

R. Grassmann, J. Burgner-Kahrs

Robotic Intelligence

🎯 What it does: A mapping method was studied to alleviate the discontinuity issues caused by branch methods (e.g., if-else statements) in continuum robots, deriving a lower triangular transformation matrix to decouple inequalities, converting interdependent inequalities into independent box constraints, and further investigating sampling, control, and workspace estimation.

On the Feasibility of EEG-based Motor Intention Detection for Real-Time Robot Assistive Control

Ho Jin Choi, Nadia Figueroa

ClassificationRobotic IntelligenceBiomedical Data

🎯 What it does: This study designs an offline data collection and training process as well as an online real-time prediction workflow, utilizing EEG signals to classify the intention of left and right arm movements, and applying the prediction results to assistive robot control;

On the Fly Robotic-Assisted Medical Instrument Planning and Execution Using Mixed Reality

Letian Ai, Alejandro Martin-Gomez

Robotic IntelligenceBiomedical Data

🎯 What it does: Propose a framework based on mixed reality that supports real-time planning and execution in robot-assisted medical systems, combining 3D anatomical image overlay, human-robot collision detection, and robot programming interfaces, along with an intuitive head-mounted display (HMD) calibration method to enhance human-robot interaction.

On the Overconfidence Problem in Semantic 3D Mapping

João Marcos Correia Marques, Kris Hauser

SegmentationDepth EstimationSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Investigated the overconfidence in fusion within semantic 3D mapping, proposed improved uncertainty calibration methods, and conducted comparisons on ScanNet.

On the Performance of Jerk-Constrained Time-Optimal Trajectory Planning for Industrial Manipulators

Jee-eun Lee, Luis Sentis

OptimizationRobotic Intelligence

🎯 What it does: Propose a jerk-constrained time-optimal trajectory planning method, which forms conservative inequality constraints by convexifying third-order constraints, and uses n-dimensional sequential linear programming (SLP) to iteratively solve for the trajectory, followed by evaluating its performance on a real robot;

On-device Self-supervised Learning of Visual Perception Tasks aboard Hardware-limited Nano-quadrotors

Elia Cereda, D. Palossi

Domain AdaptationConvolutional Neural NetworkImage

🎯 What it does: Performing self-supervised fine-tuning of pre-trained CNNs on sub-50g nano-drones to enhance performance in visual perception regression tasks.

On-the-Go Tree Detection and Geometric Traits Estimation with Ground Mobile Robots in Fruit Tree Groves

Dimitrios Chatziparaschis, Konstantinos Karydis

Object DetectionImageMultimodalityPoint CloudAgriculture Related

🎯 What it does: Propose an algorithm framework for real-time fruit tree detection and estimation of key geometric features such as width and height during vehicle movement, combining NDVI 2D images and 3D LiDAR point clouds, and achieving multi-modal feature matching and joint estimation with the Kalman filter through tree feature point association and parameter estimation algorithms.

One-vs-All Semi-Automatic Labeling Tool for Semantic Segmentation in Autonomous Driving

Jing Gu, Amine Ben Arab

SegmentationAutonomous DrivingImage

🎯 What it does: Proposed an active learning framework called OVAAL based on One-vs-All, as well as an OVA-based semi-supervised learning method called OVAAL+, which were evaluated on semantic segmentation tasks in autonomous driving scenarios.

ONeK-SLAM: A Robust Object-level Dense SLAM Based on Joint Neural Radiance Fields and Keypoints

Yue Zhuge, Zhuqing Jiang

Neural Radiance FieldSimultaneous Localization and Mapping

🎯 What it does: Proposed a robust dense SLAM system called ONeK-SLAM based on the joint use of object-level neural radiance fields and keypoints.

Online Adaptation of Sampling-Based Motion Planning with Inaccurate Models

Marco Faroni, Dmitry Berenson

OptimizationRobotic Intelligence

🎯 What it does: Proposed an online adaptive sampling-based motion planning method, utilizing model error estimation and online observations to correct the planning strategy.

Online Calibration of a Single-Track Ground Vehicle Dynamics Model by Tight Fusion with Visual-Inertial Odometry

Haolong Li, Joerg Stueckler

Autonomous DrivingOptimizationSimultaneous Localization and Mapping

🎯 What it does: Integrate a single-track dynamic model with visual-inertial odometry (VIO), and perform online calibration and model adaptation to enhance the accuracy of forward predictions for future control inputs.

Online Data-Driven Safety Certification for Systems Subject to Unknown Disturbances

Nicholas Rober, Jonathan P. How

Optimization

🎯 What it does: Developed an online data-driven security authentication method that utilizes an optimized moving horizon estimator (MHE) to estimate unknown disturbances, and integrates the estimation results into real-time reachability analysis to verify whether the system avoids hazardous states under unknown disturbances.

Online Distribution Shift Detection via Recency Prediction

Rachel Luo, M. Pavone

Anomaly DetectionRobotic IntelligenceTime SeriesSequential

🎯 What it does: Proposed an online distribution drift detection method using recency prediction technology, achieving rapid detection in high-dimensional robot data streams with a theoretical guarantee of false alarm rate < ε.

Online Estimation of Articulated Objects with Factor Graphs using Vision and Proprioceptive Sensing

Russell Buchanan, S. Vijayakumar

Pose EstimationRobotic IntelligenceSimultaneous Localization and MappingMultimodality

🎯 What it does: Propose an online estimation method that combines availability information predicted by visual neural networks with interactive kinematic sensing to identify joint objects and enable robotic autonomous opening.

Online Fault Detection in Manipulation Tasks via Generative Models

Michael W. Lanighan, Oscar Youngquist

Anomaly DetectionRobotic IntelligenceGenerative Adversarial NetworkImage

🎯 What it does: Construct an image manifold containing only correctly executed images using generative adversarial networks (GAN), and achieve online error detection in manipulation tasks by leveraging differences in image reconstruction.

Online Minimization of the Robot Silhouette Viewed From Eye-to-Hand Camera

Giovanni Cortigiani, Domenico Prattichizzo

OptimizationRobotic IntelligenceImage

🎯 What it does: A control technique was developed using redundant robots from the eye-hand camera perspective to minimize the robot's silhouette in images.

Online Multi-Contact Feedback Model Predictive Control for Interactive Robotic Tasks

Seohee Han, Min Jun Kim

Robotic Intelligence

🎯 What it does: Propose an online multi-contact feedback model predictive control (MPC) for handling multi-contact interactive robotic tasks with unknown contact positions.

Online On-Demand Multi-Robot Coverage Path Planning

Ratijit Mitra, I. Saha

Robotic IntelligenceBenchmark

🎯 What it does: Proposed an online centralized multi-robot coverage path planning algorithm based on time-domain synchronization, which recalculates paths only for robots that have fully traversed their paths on demand, while other robots reuse existing paths, significantly reducing computational burden.

Online Supervised Training of Spaceborne Vision during Proximity Operations using Adaptive Kalman Filtering

T. Park, Simone D’Amico

Pose EstimationDomain AdaptationImage

🎯 What it does: Developed an online supervised training (OST) method that utilizes an adaptive unscented Kalman filter to provide pseudo labels for a pose estimation neural network, enabling online learning with real-time aerial imagery during close-range rendezvous and proximity operations (RPO).

Online Trajectory Deformation and Tracking for Self-entanglement-free Differential-Driven Robots

Jiangpin Liu, Rong Xiong

OptimizationRobotic Intelligence

🎯 What it does: Designed and implemented an optimization-based trajectory deformation and tracking algorithm for achieving self-entanglement free (SEF) tethered differential drive robots;

Online-Learning-Based Distributionally Robust Motion Control with Collision Avoidance for Mobile Robots

Han Wang, Weidong Zhang

OptimizationRobotic Intelligence

🎯 What it does: A distributionally robust nonlinear model predictive control (NMPC) method based on online learning is proposed to achieve collision avoidance in environments with moving obstacles.

Open X-Embodiment: Robotic Learning Datasets and RT-X Models : Open X-Embodiment Collaboration0

A. Padalkar, Zichen Jeff Cui

Robotic Intelligence

🎯 What it does: A standardized dataset containing 22 robots, 527 skills, and 160,266 tasks was constructed, and a large-scale model named RT-X was trained to demonstrate positive transfer and performance improvement in general robotic policies.

Open-Fusion: Real-time Open-Vocabulary 3D Mapping and Queryable Scene Representation

Kashu Yamazaki (University of Arkansas), Ngan Le (University of Arkansas)

RecognitionSegmentationVision Language ModelMultimodalityPoint Cloud

🎯 What it does: Propose a real-time open-vocabulary 3D mapping and queryable scene representation method, using RGB-D data to achieve annotation-free 3D segmentation and open-vocabulary querying;

Open-Vocabulary Affordance Detection using Knowledge Distillation and Text-Point Correlation

Tuan V. Vo, Anh Nguyen

Knowledge DistillationPoint Cloud

🎯 What it does: Propose an open-vocabulary 3D point cloud object detection method based on knowledge distillation and text-point correlation.

OpenAnnotate3D: Open-Vocabulary Auto-Labeling System for Multi-modal 3D Data

Yijie Zhou, Wenchao Ding

Object DetectionSegmentationTransformerLarge Language ModelVision Language ModelMultimodalityChain-of-Thought

🎯 What it does: Proposed and implemented an open-source multi-modal 3D auto-annotation system named OpenAnnotate3D, which can automatically generate 2D masks, 3D masks, and 3D bounding boxes;

OpenBot-Fleet: A System for Collective Learning with Real Robots

Matthias A. M¨uller, V. Koltun

Robotic Intelligence

🎯 What it does: Introduces the OpenBot-Fleet system, which utilizes smartphone sensors, cloud storage, and low-cost wheeled robots to enable large-scale mobile robots to collect data, learn, and deploy navigation strategies in real-world environments.

Opportunistic Communication in Robot Teams

Daniel Mox, Vijay Kumar

OptimizationRobotic Intelligence

🎯 What it does: Propose a Mobile Infrastructure on Demand (MID) method that employs a dedicated team of robots to construct and maintain a wireless network, meeting the communication needs of another set of task robots in the absence of existing communication infrastructure.

Optimal Collaborative Transportation for Under-Capacitated Vehicle Routing Problems using Aerial Drone Swarms

Akash Kopparam Sreedhara, Heinz Koeppl

Optimization

🎯 What it does: Studied a model for collaborative transportation using drone swarms, defining it as a new 'under-capacity' vehicle routing problem (VRP)

Optimal Containment Control of Multiple Quadrotors via Reinforcement Learning*

Ming Cheng, Xiangke Wang

OptimizationReinforcement Learning

🎯 What it does: Explores the optimal containment control problem for nonlinear, underactuated quadrotors in multi-leader scenarios, proposing a cascade controller using reinforcement learning and deriving control protocols from historical data without requiring an exact dynamic model.

Optimal Control for Clutched-Elastic Robots: A Contact-Implicit Approach

Dennis Ossadnik, Sami Haddadin

OptimizationRobotic Intelligence

🎯 What it does: This paper proposes a contact-implicit optimization scheme that simultaneously optimizes the control inputs and clutch sequences of elastic robots to achieve precise energy transfer timing control.

Optimal Control of Granular Material

Yuichiro Aoyama, Evangelos A. Theodorou

OptimizationGraph Neural NetworkGraphPhysics Related

🎯 What it does: This paper addresses granular material systems driven by rigid bodies, using a graph neural network-based simulator to learn their dynamics, and achieving optimal control through differential dynamic programming to form target shapes.

Optimal Control Synthesis with Relaxed Global Temporal Logic Specifications for Homogeneous Multi-robot Teams

Disha Kamale, C. Vasile

OptimizationRobotic Intelligence

🎯 What it does: Control synthesis for homogeneous robot teams under given global temporal logic specifications and relaxed user preferences when feasibility fails.

Optimal Driver Warning Generation in Dynamic Driving Environment

Chenran Li, Behzad Dariush

Autonomous DrivingOptimizationReinforcement Learning

🎯 What it does: Studied the optimal driver warning generation problem in dynamic driving environments, considering the interaction between warnings, driver behavior, and vehicle states, and modeled it as a partially observable Markov decision process (POMDP)

Optimal Planning for Timed Partial Order Specifications

Kandai Watanabe, S. Sankaranarayanan

OptimizationRobotic Intelligence

🎯 What it does: Proposes a framework for multi-robot task planning aimed at minimizing the total makespan while satisfying temporal and precedence constraints.

Optimal Prescribed-Time Control based Reactive Planning System for Quadruped Robot Navigation

Shaohang Xu, Lijun Zhu

OptimizationRobotic Intelligence

🎯 What it does: Propose a reactive planning system for quadruped robots based on preset-time control, enabling omni-directional movement and correcting deviations through feedback control.

Optimal Scene Graph Planning with Large Language Model Guidance

Zhirui Dai, Nikolay Atanasov

OptimizationTransformerLarge Language ModelText

🎯 What it does: Developed an efficient task planning algorithm for hierarchical metric-semantic models, leveraging large language models (LLMs) to convert natural language tasks into LTL automata, achieving optimal hierarchical LTL planning on scene graphs.

Optimal Task Allocation for Heterogeneous Multi-robot Teams with Battery Constraints

Alvaro Calvo, Jesús Capitán

OptimizationRobotic Intelligence

🎯 What it does: Propose a task allocation method for heterogeneous multi-robot teams, considering battery constraints, task fragmentation, and synchronization in multi-robot collaboration.

Optimization and Evaluation of a Multi Robot Surface Inspection Task Through Particle Swarm Optimization

Darren Chiu, Bahar Haghighat

ClassificationOptimizationRobotic Intelligence

🎯 What it does: Studied a simplified two-class surface inspection task, using a 3 cm wheeled robot swarm to perform distributed Bayesian decision classification on a 1m×1m random black-and-white tile surface, with algorithm parameters adjusted via particle swarm optimization.

Optimization Based Dynamic Skateboarding of Quadrupedal Robot

Zhe Xu, Shiwu Zhang

OptimizationRobotic Intelligence

🎯 What it does: Developed an optimization-based control pipeline enabling the quadruped robot CyberDog2 to achieve dynamic balance and acceleration on a skateboard;

Optimization of Flexible Bronchoscopy Shape Sensing Using Fiber Optic Sensors

Xinran Liu, Hongbin Liu

OptimizationBiomedical Data

🎯 What it does: Proposed a shape assessment and optimization method for flexible bronchoscopes to handle complex spatial configurations in the human bronchial tree.

Optimized Design and Fabrication of Skeletal Muscle Actuators for Bio-syncretic Robots

Lianchao Yang, Lianqing Liu

OptimizationRobotic Intelligence

🎯 What it does: An optimized design method was proposed, combining simulation analysis with experimental preparation to fabricate multi-band structured engineered skeletal muscle tissue (eSKT), which was applied to bio-synthetic robots, achieving significant contractile force and driving functionality.

Optimizing Dynamic Balance in a Rat Robot via the Lateral Flexion of a Soft Actuated Spine

Yuhong Huang, A. Knoll

OptimizationRobotic Intelligence

🎯 What it does: Developed an optimized controller for soft elastic scoliosis to enhance dynamic balance in an experimental mouse robot during stepping locomotion, achieving balance optimization without altering leg movements.

Optimizing Modular Robot Composition: A Lexicographic Genetic Algorithm Approach

Jonathan Külz, Matthias Althoff

OptimizationRobotic Intelligence

🎯 What it does: Propose a modular robot construction optimization method combining genetic algorithms with ordinal evaluation

Optimizing Multi-Touch Textile and Tactile Skin Sensing Through Circuit Parameter Estimation

Boquan Su, Changliu Liu

Optimization

🎯 What it does: A new framework is proposed that treats the touch perception problem as a resistance sensing array estimation problem, utilizing a regularized least squares objective function to enhance touch perception accuracy and reduce ghosting effects.

OptiState: State Estimation of Legged Robots using Gated Networks with Transformer-based Vision and Kalman Filtering

Alexander Schperberg, Dennis Hong

OptimizationRobotic IntelligenceRecurrent Neural NetworkTransformerAuto EncoderSimultaneous Localization and MappingImage

🎯 What it does: A hybrid method is proposed, combining Kalman filtering, convex MPC optimization, and learning techniques. It utilizes joint encoders, IMU, and ground reaction force control outputs, and employs GRU gated networks and Vision Transformer autoencoders to perform semantic and height reasoning on depth images, achieving precise estimation of the quadruped robot's trunk state.

Orbit-Surgical: An Open-Simulation Framework for Learning Surgical Augmented Dexterity

Qinxi Yu, Animesh Garg

Robotic IntelligenceReinforcement LearningBenchmark

🎯 What it does: This paper proposes Orbit-Surgical, a physics-based surgical robot simulation framework with realistic rendering from NVIDIA Omniverse, providing 14 benchmark surgical tasks for da Vinci Research Kit (dVRK) and Smart Tissue Autonomous Robot (STAR), and leveraging GPU parallelization to train reinforcement learning and imitation learning algorithms, enabling robot learning to assist human surgical skills, and demonstrating policy transfer from simulation to real-world environments.

Orientation-Aware Multi-Modal Learning for Road Intersection Identification and Mapping

Qibin He, Li Sun

Object DetectionAutonomous DrivingMultimodalityPoint CloudSequential

🎯 What it does: Proposed a direction-oriented multi-modal learning paradigm, formulating the intersection recognition task as a directed object detection problem.

OriTrack: A Small, 3 Degree-of-Freedom, Origami Solar Tracker

Crystal E. Winston, L. Casey

🎯 What it does: Design, manufacture, and control a small 3-degree-of-freedom origami solar tracker named OriTrack

OSCaR: An Origami-Inspired Shape-Changing Robot for Ground Coverage Tasks

Zirui Fan, Hongying Zhang

Robotic Intelligence

🎯 What it does: Developed a foldable and deployable robot called OSCaR to enhance adaptability in ground coverage tasks such as floor cleaning.

Osiris: Building Hierarchical Representations for Agricultural Environments

Adam Mukuddem, Paul Amayo

Representation LearningPoint CloudGraphAgriculture Related

🎯 What it does: Built a 3D scene graph generation system called Osiris for agricultural environments, capable of incrementally generating hierarchical graph representations from data collected by mobile robots.

Ospreys-inspired Self-takeoff Strategy of An Eagle-scale Flapping-wing Robot: System Design and Flight Experiments

Haoyu Wang, Erzhen Pan

Robotic Intelligence

🎯 What it does: First achievement of autonomous takeoff for a hawk-level flapping-wing robot, proposing a bio-inspired takeoff strategy inspired by pelicans, and designing a system comprising a flapping-wing robot and an auxiliary platform, conducting flight experiments under different thrust-to-weight ratios and takeoff angles.

OSSAR: Towards Open-Set Surgical Activity Recognition in Robot-assisted Surgery

Long Bai, Hongliang Ren

RecognitionRobotic IntelligenceVideoBiomedical Data

🎯 What it does: Proposed an open-set surgical activity recognition framework (OSSAR) for robot-assisted surgery

Out of Sight, Still in Mind: Reasoning and Planning about Unobserved Objects with Video Tracking Enabled Memory Models

Yixuan Huang, Tucker Hermans

Object TrackingRobotic IntelligenceTransformerWorld ModelPoint Cloud

🎯 What it does: Proposed two models, DOOM and LOOM, leveraging transformer relational dynamics and video tracking enhanced memory models to encode memory for occluded objects in multi-object manipulation by robots, supporting reasoning and planning.

Output-Sampled Model Predictive Path Integral Control (o-MPPI) for Increased Efficiency

Leon Yan, S. Devasia

Autonomous DrivingOptimizationReinforcement Learning

🎯 What it does: Developed o-MPPI, an output-sampling-based MPPI, to improve sampling efficiency in meeting output constraints under dynamic environments

Outram: One-shot Global Localization via Triangulated Scene Graph and Global Outlier Pruning

Peng Yin, Lihua Xie

Pose EstimationRetrievalAutonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a hierarchical one-time LiDAR localization algorithm called Outram, which utilizes substructures of 3D scene graphs for local consistency correspondence search and global substructure-level outlier removal, combining feature retrieval and correspondence extraction to achieve consistency refinement from local to global, addressing substructure ambiguity issues.

Overcoming Hand and Arm Occlusion in Human-to-Robot Handovers: Predicting Safe Poses with a Multimodal DNN Regression Model

C. Lollett, Shigeki Sugano

Pose EstimationRobotic IntelligenceMultimodality

🎯 What it does: A three-branch multimodal DNN regression model was developed to predict safe robotic arm poses under occlusion conditions.

Overparametrization helps offline-to-online generalization of closed-loop control from pixels

Mathias Lechner, Daniela Rus

Robotic IntelligenceConvolutional Neural NetworkTransformerReinforcement LearningImage

🎯 What it does: Investigate the generalization ability of visual models from open training to closed-loop deployment, evaluating their closed-loop test performance in both in-distribution and out-of-distribution scenarios.

PanNote: an Automatic Tool for Panoramic Image Annotation of People’s Positions

Alberto Bacchin, Emanuele Menegatti

Object DetectionConvolutional Neural NetworkImagePoint Cloud

🎯 What it does: Developed and validated a tool called PanNote for automatic annotation of human positions in panoramic videos

Parallel Optimization with Hard Safety Constraints for Cooperative Planning of Connected Autonomous Vehicles

Zhen Huang, Jun Ma

Autonomous DrivingOptimization

🎯 What it does: Formulate cooperative autonomous driving tasks as an optimal control problem with hard safety constraints, propose a computationally efficient parallel optimization framework, maintain problem convexity through convex approximation, and use iterative nearest neighbor search to determine the optimal passage order;

Parameter-efficient Prompt Learning for 3D Point Cloud Understanding

Hongyu Sun, Deying Li

ClassificationSegmentationComputational EfficiencyTransformerPrompt EngineeringPoint Cloud

🎯 What it does: Proposes a parameter-efficient Prompt tuning method called PPT to adapt large multimodal models for 3D point cloud understanding tasks.

Partial Belief Space Planning for Scaling Stochastic Dynamic Games

K. Vakil, Alyssa Pierson

Computational EfficiencyReinforcement Learning

🎯 What it does: Propose a method that reduces the computational burden in stochastic dynamic games by utilizing partial belief propagation

Particle Filter with Stable Embedding for State Estimation of the Rigid Body Attitude System on the Set of Unit Quaternions

Hee-Deok Jang, D. Chang

Pose EstimationComputational Efficiency

🎯 What it does: Proposes a particle filter (Particle Filter) combined with a stable embedding method for state estimation of rigid body attitude systems on the unit quaternion (S³) manifold.

Passive Aligning Physical Interaction of Fully-Actuated Aerial Vehicles for Pushing Tasks

Tong Hui, R. Siegwart

Robotic Intelligence

🎯 What it does: Proposed a control strategy for fully actuated UAVs to achieve passive alignment during pushing tasks

PathRL: An End-to-End Path Generation Method for Collision Avoidance via Deep Reinforcement Learning

Wenhao Yu, Jianmin Ji

Autonomous DrivingReinforcement Learning

🎯 What it does: Proposes PathRL, a method that directly generates navigation paths through deep reinforcement learning by training strategies.

Pay Attention to How You Drive: Safe and Adaptive Model-Based Reinforcement Learning for Off-Road Driving

Sean J. Wang, Aaron M. Johnson

Autonomous DrivingTransformerReinforcement Learning

🎯 What it does: This paper proposes a model-based reinforcement learning method for off-road driving, training a System Identification Transformer (SIT) and an Adaptive Dynamics Model (ADM), and using an online risk-aware model with a path integral controller to achieve safe control.

PBP: Path-based Trajectory Prediction for Autonomous Driving

S. Afshar, Henggang Cui

Autonomous DrivingSequential

🎯 What it does: Proposed and implemented a path-based trajectory prediction model (PBP), which performs discrete probability prediction on the reference path distribution in HD maps and generates trajectories in the relative Frenet coordinate system;

PCB-RandNet: Rethinking Random Sampling for LiDAR Semantic Segmentation in Autonomous Driving Scene

Hui Cheng, Guoqiang Xiao

SegmentationAutonomous DrivingPoint Cloud

🎯 What it does: Proposes Polar Cylinder Balanced Random Sampling (PCB-RandNet) and introduces a sampling consistency loss for semantic segmentation of large-scale LiDAR point clouds.

Pedestrian Trajectory Prediction Using Dynamics-based Deep Learning

Honghui Wang, Rohitash Chandra

Object TrackingTransformerSequentialBenchmarkOrdinary Differential Equation

🎯 What it does: Proposed a dynamics-based deep learning framework that integrates an asymptotically stable dynamics system into the Transformer model for pedestrian trajectory prediction.

Pedipulate: Enabling Manipulation Skills using a Quadruped Robot’s Leg

Philip Arm, Marco Hutter

Robotic IntelligenceReinforcement Learning

🎯 What it does: Achieve leg-based single-foot position tracking through training reinforcement learning strategies, developing a walkable-manipulable controller capable of performing manipulation tasks within a large workspace. Real-world experiments on quadruped robots demonstrate successful completion of tasks such as door opening, sampling, and obstacle pushing, with the ability to carry loads exceeding 2.0 kg.

Pegasus: a Novel Bio-inspired Quadruped Robot with Underactuated Wheeled-Legged Mechanism *

Yuzhen Pan, Huiliang Shang

OptimizationRobotic Intelligence

🎯 What it does: Designed and analyzed the Pegasus quadruped wheeled robot, providing two motion modes and achieving the ability to adapt to different tasks.

PeLiCal: Targetless Extrinsic Calibration via Penetrating Lines for RGB-D Cameras with Limited Co-visibility

Jaeho Shin, Ayoung Kim

Pose EstimationDepth EstimationImage

🎯 What it does: Proposes PeLiCal, a target-free, real-time, and robust RGB-D camera extrinsic calibration method based on long straight line features in the scene.

Perception through Cognitive Emulation : "A Second Iteration of NaivPhys4RP for Learningless and Safe Recognition and 6D-Pose Estimation of (Transparent) Objects"

Franklin Kenghagho Kenfack, Michael Beetz

RecognitionPose EstimationSafty and PrivacyPhysics Related

🎯 What it does: Designed and implemented the first version of NaivPhys4RP, which can achieve learning-free and risk-free object recognition and 6D pose estimation under challenging sensor conditions such as data scarcity, occlusion, transparency, low-quality depth, or handheld scenarios.

Perception-and-Energy-aware Motion Planning for UAV using Learning-based Model under Heteroscedastic Uncertainty

Reiya Takemura, G. Ishigami

OptimizationRobotic IntelligencePoint Cloud

🎯 What it does: Proposes a UAV motion planning method under GNSS-restricted environments, combining energy consumption and LiDAR perception quality, achieving efficient and reliable flight through cost function optimization.

Perceptual Factors for Environmental Modeling in Robotic Active Perception

David Morilla-Cabello, Eduardo Montijano

Robotic IntelligenceGraph Neural NetworkSimultaneous Localization and MappingWorld ModelImage

🎯 What it does: Proposed a novel active perception framework that integrates perceptual factors into planning and fusion, quantifying the environmental impact on candidate perspectives through constructing a perceptual graph, and adjusting measurement uncertainty accordingly.

Persistent Monitoring of Large Environments with Robot Deployment Scheduling in between Remote Sensing Cycles

Kizito Masaba, Alberto Quattrini Li

OptimizationRobotic Intelligence

🎯 What it does: Studied the deployment planning of autonomous surface vehicles (ASVs) based on prior data for time and location to continuously monitor spatiotemporal phenomena in environments such as large lakes.

Personality- and Memory-Based Software Framework for Human-Robot Interaction

Alice Nardelli, C. Recchiuto

Robotic Intelligence

🎯 What it does: Proposed a psychological cognitive architecture based on robot personality and memory for human-robot interaction

Phase Synthesis for Spatial Locomotion Control of Retractable Worm Robots

Zhongcheng Wang, Bin Liang

Robotic Intelligence

🎯 What it does: A phase synthesis scheme is proposed for motion control of a scalable worm robot in three-dimensional space, and based on this scheme, the RW-Robot prototype was developed. The foot phase is defined, motion is divided into short-term and long-term, and a frequency domain compression mode is applied, ultimately generating a new spatial gait. The effectiveness of the scheme in turning and climbing was verified through actual experiments.

Phasic Diversity Optimization for Population-Based Reinforcement Learning

Jing Jiang, Xin Yang

OptimizationReinforcement Learning

🎯 What it does: Proposed and implemented the Phasic Diversity Optimization (PDO) algorithm, which separates different stages in reward and diversity training, and employs diversity enhancement in the auxiliary phase without affecting performance.

Physical and Digital Adversarial Attacks on Grasp Quality Networks

N. Alharthi, Martim Brandão

Adversarial AttackRobotic Intelligence

🎯 What it does: Proposed and experimentally tested two physical and digital adversarial attacks targeting grasp quality networks

Physical Priors Augmented Event-Based 3D Reconstruction

Jiaxu Wang, Renjing Xu

GenerationNeural Radiance FieldPhysics Related

🎯 What it does: Propose to enhance NeRF training by leveraging motion, geometry, and density priors, and introduce a density-guided patch sampling strategy that enables direct 3D scene reconstruction from event streams, while constructing the first large-scale event-based 3D reconstruction dataset.

Physically Grounded Vision-Language Models for Robotic Manipulation

Jensen Gao, Dorsa Sadigh

Robotic IntelligenceLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality

🎯 What it does: Create the PHYSOBJECTS dataset, fine-tune a vision-language model using it, integrate it with a robot planner based on a large language model, and evaluate on both simulated and real robots.

Physics-Encoded Graph Neural Networks for Deformation Prediction under Contact

Mahdi Saleh, Federico Tombari

Graph Neural NetworkGraphPhysics Related

🎯 What it does: Using Physics-Encoded Graph Neural Networks to predict deformation under contact between rigid and deformable bodies.

Physics-informed Neural Motion Planning on Constraint Manifolds

Ruiqi Ni, A. H. Qureshi

OptimizationComputational EfficiencyRobotic IntelligencePhysics Related

🎯 What it does: Proposed and implemented the first physics-informed CMP framework for solving the Eikonal equation on constrained manifolds, which uses neural networks to train motion planning functions without requiring expert data.

Physics-Informed Neural Network for Multirotor Slung Load Systems Modeling

Gil Serrano, Rita Cunha

Robotic IntelligenceRecurrent Neural NetworkPhysics Related

🎯 What it does: Using physics-informed neural networks for end-to-end modeling of a multirotor suspended load system and predicting future state sequences.

Physics-Informed Neural Networks for Continuum Robots: Towards Fast Approximation of Static Cosserat Rod Theory

Martin Bensch, Moritz Schappler

Computational EfficiencyRobotic IntelligencePhysics Related

🎯 What it does: Compute the overall shape of tension-driven continuum robots using physics-informed neural networks (PINN) based on Cosserat rod theory.

PillarGen: Enhancing Radar Point Cloud Density and Quality via Pillar-based Point Generation Network

Ji Song Kim, J. Choi

Object DetectionGenerationData SynthesisAutonomous DrivingPoint Cloud

🎯 What it does: Transfer radar point clouds from one domain to another using the PillarGen model, generating denser and higher quality point clouds.

Pilot comparison of customized and generalized hip-knee-ankle exoskeleton torque profiles

Gwendolyn M. Bryan, S. Collins

OptimizationRobotic Intelligence

🎯 What it does: Compare the effects of personalized, other, and average torque profiles on metabolic cost using previously optimized hip-knee-ankle torque profiles in three expert users.

PlaceNav: Topological Navigation through Place Recognition

Lauri Suomela, Joni-Kristian Kämäräinen

RecognitionRetrievalRobotic IntelligenceImage

🎯 What it does: Propose PlaceNav, which decomposes robot-agnostic components into navigation-specific and general computer vision modules, leveraging visual place recognition for subgoal selection and Bayesian filtering to enhance subgoal temporal consistency.