arXivSub Start free trial

ICRA 2024 Papers — Page 13

IEEE International Conference on Robotics and Automation · 1760 papers

PlanCollabNL: Leveraging Large Language Models for Adaptive Plan Generation in Human-Robot Collaboration

Silvia Izquierdo-Badiola, Guillem Alenyà

Robotic IntelligenceTransformerLarge Language ModelText

🎯 What it does: Proposes the PlanCollabNL framework, which leverages LLM to convert natural language goals and agent conditions into structured task planning problems, and generates subgoals and allocation suggestions.

Planning of Explanations for Robot Navigation

Amar Halilovic, Senka Krivic

Explainability and InterpretabilityRobotic IntelligenceMultimodality

🎯 What it does: This paper models robot navigation explanation generation as an automated planning problem, generating visual and textual explanations influenced by the robot's personality while considering context, environmental, and spatial features; subsequent user studies compare multimodal and unimodal explanations, revealing that multimodal explanations are more satisfactory to users, while explanations from robots with extreme personalities receive lower satisfaction.

Planning Optimal Trajectories for Mobile Manipulators under End-effector Trajectory Continuity Constraint

Quang-Nam Nguyen, Quang Pham

OptimizationRobotic Intelligence

🎯 What it does: This paper proposes a motion planning method for mobile manipulators under the constraint of end-effector trajectory continuity. It first plans an optimal base trajectory that satisfies geometric task constraints, collision avoidance, and base speed constraints, and then calculates the manipulator trajectory based on this trajectory. Additionally, a discrete optimal base trajectory planning algorithm is provided for mobile printing tasks.

Planning with Learned Subgoals Selected by Temporal Information

Xi Huang, Torsten Kröger

Autonomous Driving

🎯 What it does: Proposed a method that utilizes a generative model to progressively generate sub-goals and filters these sub-goals using temporal information and a learned time estimator, thereby achieving path planning in dynamic environments.

Pluck and Play: Self-supervised Exploration of Chordophones for Robotic Playing

Michael Görner, Jianwei Zhang

Robotic IntelligenceMultimodalityAudio

🎯 What it does: Through audio-tactile exploration, a physical robot arm directly characterizes the geometric model and audio start response of string instruments (e.g., guzheng). It first performs multiple plucking refinements using previously estimated string positions (provided by tactile teaching or visual estimation), then employs a Gaussian process-based safe active exploration paradigm to explore feasible plucking actions' audio start responses. The generated model enables imprecise robot arms to play note sequences with varying loudness on the guzheng.

Plug in the Safety Chip: Enforcing Constraints for LLM-driven Robot Agents

Ziyi Yang, Stefanie Tellex

Safty and PrivacyRobotic IntelligenceLarge Language ModelText

🎯 What it does: Propose a queryable safety constraint module based on Linear Temporal Logic (LTL) to ensure LLM-driven robot agents comply with safety constraints during task execution;

Pneumatic Back Exoskeleton for Lifting Posture Detection and Correction

Yu Chen, Yifan Wang

Pose EstimationTime Series

🎯 What it does: Designed and tested a pneumatic back exoskeleton for enhancing posture detection and correction

POAQL: A Partially Observable Altruistic Q-Learning Method for Cooperative Multi-Agent Reinforcement Learning

Lesong Tao, Nanning Zheng

Reinforcement Learning

🎯 What it does: Proposed a partially observable altruistic Q-learning method (POAQL) for multi-agent path finding (MAPF), addressing the misleading effects of traditional team rewards in partially observable environments by considering the cumulative rewards of observed sub-teams, and designed a guidance-free conflict resolution mechanism to emphasize cooperation.

POE: Acoustic Soft Robotic Proprioception for Omnidirectional End-effectors

Uksang Yoo, Jean Oh

Robotic IntelligenceMeshAudio

🎯 What it does: Developed a tendon-driven soft robotic hand with embedded microphones called POE, and proposed the POE-M framework, achieving shape self-sensing and high-resolution mesh reconstruction of the soft robotic hand through acoustic signals.

Point Cloud Models Improve Visual Robustness in Robotic Learners

Skand Peri, Stefan Lee

Robotic IntelligenceReinforcement LearningWorld ModelPoint Cloud

🎯 What it does: This paper studies the robustness of RGB-D and point cloud visual control strategies under different visual conditions, and proposes the Point Cloud World Model (PCWM) and point cloud-based control strategies.

Point Cloud-Based Control Barrier Function Regression for Safe and Efficient Vision-Based Control

Massimiliano de Sa, K. Sreenath

OptimizationSafty and PrivacyPoint Cloud

🎯 What it does: Synthesize low-computation-cost control barrier functions (CBF) for visual control using point cloud data, and apply them to CBF-QP to achieve safe navigation.

Point-Wise Vibration Pattern Production via a Sparse Actuator Array for Surface Tactile Feedback

Xiaosa Li, Wenbo Ding

Optimization

🎯 What it does: A point vibration pattern generation at arbitrary contact points was achieved on a tactile feedback panel with the same size as a smartphone using only five sparsely placed passive coil actuators.

PointSSC: A Cooperative Vehicle-Infrastructure Point Cloud Benchmark for Semantic Scene Completion

Yuxiang Yan, Jian Pu

SegmentationAutonomous DrivingTransformerVision Language ModelPoint CloudBenchmark

🎯 What it does: Proposed PointSSC, the first cooperative vehicle-infrastructure point cloud semantic scene completion benchmark, and developed an automatic annotation pipeline based on Semantic Segment Anything; simultaneously proposed a LiDAR-based model that combines Spatial-Aware Transformer for global and local feature extraction, and introduces a Completion and Segmentation Cooperative Module for joint completion and segmentation.

Policy Optimization by Looking Ahead for Model-based Offline Reinforcement Learning

Yang Liu, Marius Hofert

Reinforcement LearningWorld ModelBenchmark

🎯 What it does: Propose a model-based offline reinforcement learning method called POLA, which predicts future states and jointly optimizes the policy on current and future states to better maximize cumulative rewards.

POLITE: Preferences Combined with Highlights in Reinforcement Learning

Simon Holk, Iolanda Leite

Reinforcement Learning

🎯 What it does: Propose a method that integrates path preferences with high-information state-action pairs (highlighted) in reinforcement learning to reduce the required amount of preference annotations.

PoseFusion: Multi-Scale Keypoint Correspondence for Monocular Camera-to-Robot Pose Estimation in Robotic Manipulation

Xujun Han, Zheng Kan

Pose EstimationRobotic IntelligenceConvolutional Neural NetworkImage

🎯 What it does: Designed an end-to-end pose estimation method based on monocular images to achieve calibration from the camera to the robot's pose.

PPO-Based Dynamic Control of Uncertain Floating Platforms in Zero-G Environment

Mahya Ramezani (University of Luxembourg), A. Hein (University of Luxembourg)

OptimizationReinforcement LearningPhysics Related

🎯 What it does: Proposed and implemented a dynamic control method combining Proximal Policy Optimization (PPO) with Model Predictive Control (MPC) for floating platforms in zero-gravity environments, and conducted simulation and experimental validation in the Zero-G laboratory at the University of Luxembourg.

Practical and Safe Navigation Function Based Motion Planning of UAVs

Himani Sinhmar, S. Cairano

Autonomous DrivingOptimization

🎯 What it does: A safety motion planning method for unmanned aerial vehicles (UAVs) based on Explicit Reference Governor (ERG) is proposed, which synthesizes a Lyapunov function using a small amount of experimental data and modeling error assumptions. An ERG is constructed to modify flight target points in real-time, ensuring safe operation under limited power and computational resources.

Pre-Trained Masked Image Model for Mobile Robot Navigation

Vishnu Dutt Sharma, Pratap Tokekar

Robotic IntelligenceAuto EncoderImage

🎯 What it does: Utilizing pre-trained Masked Autoencoder (MAE) to achieve structure prediction-driven tasks such as field of view expansion, single-agent topological exploration, and multi-agent indoor mapping in mobile robot navigation;

Predicting against the Flow: Boosting Source Localization by Means of Field Belief Modeling using Upstream Source Proximity

F. Busch, Daniel A. Duecker

Robotic IntelligencePhysics Related

🎯 What it does: Propose a method that combines source localization with Gaussian Markov Random Field (GMRF), enhancing source localization accuracy through real-time updates of concentration and flow field beliefs, and introducing the Upstream Source Proximity (USP) metric to predict the source location.

Predicting the Intention to Interact with a Service Robot: the Role of Gaze Cues

Simone Arreghini, Antonio Paolillo

ClassificationRobotic Intelligence

🎯 What it does: Studied how to predict the interaction intentions of service robot users through visual cues

Prediction of pose errors implied by external forces applied on robots: towards a metric for the control of collaborative robots

Vincent Fortineau, D. Daney

Robotic Intelligence

🎯 What it does: Developed a method to quantify the pose deviation of robots under external disturbances, validated through simulation and real robot experiments;

Preliminary Study of Fingertip and Wrist Motion Based Haptic Controller for Robotically Assisted Micro- and Supermicrosurgery

Muneaki Miyasaka, Kotaro Tadano

Robotic Intelligence

🎯 What it does: Developed a tactile controller prototype based on fingertip and wrist movements, and evaluated its accuracy and workspace in simulation.

Preprocessing-based Kinodynamic Motion Planning Framework for Intercepting Projectiles using a Robot Manipulator

Ramkumar Natarajan, M. Likhachev

Robotic IntelligenceImage

🎯 What it does: Proposed and implemented a preprocessing-based dynamic motion planning framework for robotic manipulators to intercept flying projectiles, and built an end-to-end perception, prediction, and execution pipeline.

Privacy Risks in Reinforcement Learning for Household Robots

Miao Li, Ding Zhao

Safty and PrivacyAdversarial AttackRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a gradient inversion-based attack method that targets the training process of value-based and gradient-based reinforcement learning algorithms to reconstruct states, actions, and supervision signals.

Proactive Robot Control for Collaborative Manipulation Using Human Intent

Zhanibek Rysbek, Milos Zefran

Machine LearningRobotic Intelligence

🎯 What it does: Proposed a hierarchical robot control framework for interaction with human collaborators in collaborative operations through real-time intent recognition, implemented and validated with human studies on the UR10e robot.

Probabilistic 3D Multi-Object Cooperative Tracking for Autonomous Driving via Differentiable Multi-Sensor Kalman Filter

H. Chiu, Stephen F. Smith

Object TrackingAutonomous DrivingMultimodality

🎯 What it does: Propose a three-dimensional multi-target cooperative tracking algorithm based on differentiable multi-sensor Kalman filter.

Probabilistic Active Loop Closure for Autonomous Exploration

He Yin, Richard Kim

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed a probabilistic active loop closure framework to maximize the reduction of pose graph uncertainty and enhance the quality of occupancy maps during the autonomous exploration of mobile robots.

Probabilistic Motion Planning and Prediction via Partitioned Scenario Replay

O. D. Groot, Laura Ferranti

Autonomous DrivingOptimizationSafty and Privacy

🎯 What it does: Proposes a joint prediction and planning framework called Partitioned Scenario Replay (PSR), which achieves probabilistic collision avoidance based on real data by storing and partitioning historically observed human trajectories, and replaying scenarios similar to the current context as motion predictions during planning;

Probabilistic Spiking Neural Network for Robotic Tactile Continual Learning

Senlin Fang, Xinyu Wu

Robotic IntelligenceSpiking Neural Network

🎯 What it does: Propose a probabilistic spiking neural network framework PSNN-VCL for robotic tactile continual learning to address the catastrophic forgetting problem in ANNs.

Probable Object Location (POLo) Score Estimation for Efficient Object Goal Navigation

Jiaming Wang, Harold Soh

Computational EfficiencyReinforcement LearningBenchmark

🎯 What it does: Propose a framework based on the Probable Object Location (POLo) score, and use the POLoNet neural network to approximate the POLo score for efficient object search.

Procedure Recognition by Knowledge-Driven Segmentation in Robotic-Assisted Vitreoretinal Surgery

Zhen Li, Guibin Bian

RecognitionRobotic IntelligenceMultimodalityBiomedical Data

🎯 What it does: Proposed a multi-modal, domain knowledge-driven segmentation method called MSPR-DKS for identifying the ILM peeling process in robot-assisted retinal surgery, and constructed a corresponding dataset.

ProEqBEV: Product Group Equivariant BEV Network for 3D Object Detection in Road Scenes of Autonomous Driving

Hongwei Liu, Xian Wei

Object DetectionAutonomous DrivingImagePoint Cloud

🎯 What it does: Propose a BEV network called ProEqBEV based on equivariance to product groups for 3D object detection under multi-sensor fusion.

PROGrasp: Pragmatic Human-Robot Communication for Object Grasping

Gi-Cheon Kang, Byoung-Tak Zhang

Robotic IntelligenceVision-Language-Action ModelMultimodality

🎯 What it does: Proposed a new interactive object grasping task called Pragmatic-IOG, created the corresponding dataset IM-Dial, and designed a system named PROGrasp capable of interpreting user intent and completing target object recognition and grasping.

Projected Task-Specific Layers for Multi-Task Reinforcement Learning

J. S. Roberts, Julia Di

Reinforcement Learning

🎯 What it does: Proposed a new multi-task reinforcement learning architecture called Projected Task-Specific Layers (PTSL).

Projection-Based Fast and Safe Policy Optimization for Reinforcement Learning

Shijun Lin, Zheng Kan

Safty and PrivacyReinforcement Learning

🎯 What it does: Proposed a fast and safe policy optimization algorithm called FSPO, consisting of three steps: reward improvement update, projection to the neighborhood of the baseline policy, and projection back to the constraint set.

Prompt, Plan, Perform: LLM-based Humanoid Control via Quantized Imitation Learning

Jingkai Sun, Renjing Xu

Robotic IntelligenceTransformerLarge Language Model

🎯 What it does: Proposes a framework that combines adversarial imitation learning with large language models (LLMs), utilizing LLMs as policy planners to achieve zero-shot task control of humanoid robots using a single policy network.

Prompting Multi-Modal Tokens to Enhance End-to-End Autonomous Driving Imitation Learning with LLMs

Yiqun Duan, Renjing Xu

Autonomous DrivingTransformerLarge Language ModelPrompt EngineeringImageMultimodalityPoint Cloud

🎯 What it does: Propose a hybrid end-to-end learning framework that integrates visual and LiDAR multimodal prompts, combining basic driving imitation learning with large language models (LLM) for autonomous driving.

Prosthetic Upper-Limb Sensory Enhancement (PULSE): a Dual Haptic Feedback Device in a Prosthetic Socket

A. Ivani, A. Bicchi

Biomedical Data

🎯 What it does: This study developed and validated a dual-mode tactile feedback device called PULSE fully integrated into a prosthetic sleeve, and evaluated its effectiveness through object discrimination tasks.

Pseudo Labeling and Contextual Curriculum Learning for Online Grasp Learning in Robotic Bin Picking

Huy Le, Ngo Anh Vien

Robotic IntelligenceConvolutional Neural NetworkReinforcement Learning

🎯 What it does: Proposed an online grasping learning method SSL-ConvSAC that combines semi-supervised learning and reinforcement learning for robotic box picking tasks.

Pulsating Fluidic Sensor for Sensing of Location, Pressure and Contact Area

Joanna Jones, D. D. Damian

🎯 What it does: Designed and implemented a pulsed fluidic soft sensor capable of detecting the position, pressure, and contact area of pressing events.

PUMA: Fully Decentralized Uncertainty-aware Multiagent Trajectory Planner with Real-time Image Segmentation-based Frame Alignment

Kota Kondo, Jonathan P. How

Autonomous DrivingOptimizationSimultaneous Localization and MappingImage

🎯 What it does: Proposed an uncertainty-aware multi-agent trajectory planner and a frame alignment pipeline based on real-time image segmentation to enable safe navigation in fully decentralized environments.

PVTransformer: Point-to-Voxel Transformer for Scalable 3D Object Detection

Zhaoqi Leng, Mingxing Tan

Object DetectionAutonomous DrivingTransformerPoint Cloud

🎯 What it does: Propose PVTransformer, which uses a Transformer attention module to replace the traditional PointNet, mapping sparse point clouds to voxels;

QuAD: Query-based Interpretable Neural Motion Planning for Autonomous Driving

Sourav Biswas, R. Urtasun

Autonomous DrivingExplainability and InterpretabilityComputational Efficiency

🎯 What it does: Proposes a query-based interpretable neural motion planning framework that directly queries occupancy information at relevant spatiotemporal points to evaluate candidate trajectories.

Quadcopter Trajectory Time Minimization and Robust Collision Avoidance via Optimal Time Allocation

Zhefan Xu, Kenji Shimada

Autonomous DrivingOptimizationRobotic Intelligence

🎯 What it does: Propose the Robust Optimal Time Allocation (ROTA) framework as a trajectory post-processing tool, enhancing trajectory execution time efficiency and safety under uncertain conditions through time progression optimization.

Quadratic Programming Based Inverse Kinematics for Precise Bimanual Manipulation

T. Chaki, Tomohiro Kawakami

OptimizationRobotic Intelligence

🎯 What it does: A quadratic programming-based inverse kinematics method is proposed for precise motion in dual-arm collaboration, which ensures the relative position of end-effectors remains within joint limit and task space reachability constraints by incorporating maximum tolerable error as inequality constraints; meanwhile, virtual springs are introduced to achieve fine force collaboration.

QuadricsNet: Learning Concise Representation for Geometric Primitives in Point Clouds

Ji Wu, Gui-Song Xia

Representation LearningPoint Cloud

🎯 What it does: Proposed a new framework that represents 3D point cloud geometry primitives using quadratic surfaces and implemented an end-to-end learning model called QuadricsNet to parse quadratic surfaces.

Quadruped-Frog: Rapid Online Optimization of Continuous Quadruped Jumping

Guillaume Bellegarda, A. Ijspeert

OptimizationRobotic Intelligence

🎯 What it does: Rapid online optimization of quadruped robot jumping control on hardware, designing foot mechanics curves with only a few parameters and directly optimizing them via Bayesian optimization; subsequently tracking these mechanics curves at the joint level, and combining them with Cartesian PD damping control and virtual model control to achieve multi-directional (forward, lateral, rotational) jumping;

Quantized Visual-Inertial Odometry

Yuxiang Peng, Guoquan Huang

Pose EstimationComputational EfficiencySimultaneous Localization and MappingMultimodality

🎯 What it does: Propose Quantized Visual Inertial Odometry (QVIO), which significantly reduces data transmission by quantizing the original visual measurements and using them in EKF updates or MAP estimation.

Quasi-static Path Planning for Continuum Robots By Sampling on Implicit Manifold

Yifan Wang, Yue Chen

Robotic Intelligence

🎯 What it does: Proposing a method for quasi-static path planning on implicit manifolds

Quasi-static Soft Fixture Analysis of Rigid and Deformable Objects

Yifei Dong, Florian T. Pokorny

OptimizationComputational EfficiencyRobotic IntelligencePhysics Related

🎯 What it does: Proposes a sampling-based method for studying capture operations under energy constraints for rigid and simplified deformable 3D objects, and introduces the concept of soft grippers.

QUEST: Query Stream for Practical Cooperative Perception

Siqi Fan, Zaiqing Nie

Autonomous DrivingExplainability and InterpretabilitySequential

🎯 What it does: Propose the concept of query collaboration and build the QUEST framework, supporting query flow interaction between agents.

R-LGP: A Reachability-guided Logic-geometric Programming Framework for Optimal Task and Motion Planning on Mobile Manipulators

K. Ly, Ioannis Havoutis

OptimizationRobotic Intelligence

🎯 What it does: Propose an optimization-based mobile manipulator task and motion planning framework.

Radar Tracker: Moving Instance Tracking in Sparse and Noisy Radar Point Clouds

Matthias Zeller, C. Stachniss

Object TrackingTransformerPoint CloudBenchmark

🎯 What it does: Propose a learning-based radar tracker that tracks moving instances in sparse radar point clouds, achieving association through the integration of temporal offset predictions, attention mechanisms, and geometric-appearance features.

Radar-Only Odometry and Mapping for Autonomous Vehicles

Daniel Casado Herraez, C. Stachniss

Autonomous DrivingSimultaneous Localization and MappingPoint Cloud

🎯 What it does: Proposes a method for pose estimation and mapping using only radar, including two odometry estimation schemes and a map filtering step

RadarCam-Depth: Radar-Camera Fusion for Depth Estimation with Learned Metric Scale

Han Li, Xingxing Zuo

Depth EstimationImagePoint Cloud

🎯 What it does: Proposes a method that fuses single-view images with sparse noisy radar point clouds to achieve metric dense depth estimation.

RadCloud: Real-Time High-Resolution Point Cloud Generation Using Low-Cost Radars for Aerial and Ground Vehicles

David Hunt, Miroslav Pajic

GenerationAutonomous DrivingPoint Cloud

🎯 What it does: Propose RadCloud, a real-time framework that can directly generate high-resolution LiDAR-like 2D point clouds from low-resolution radar frames for resource-constrained drones and ground vehicles.

RainbowSight: A Family of Generalizable, Curved, Camera-Based Tactile Sensors For Shape Reconstruction

Megha H. Tippur, Edward H. Adelson

Depth EstimationRobotic IntelligenceImage

🎯 What it does: Design and propose the RainbowSight curved camera-based tactile sensor, employing addressable RGB LED rainbow spectral illumination and improved calibration methods to achieve high-resolution depth reconstruction.

RaLF: Flow-based Global and Metric Radar Localization in LiDAR Maps

Abhijeet Nayak, Abhinav Valada

Pose EstimationAutonomous DrivingFlow-based ModelMultimodalityPoint Cloud

🎯 What it does: Proposes RaLF, a method based on deep neural networks for radar scan localization in LiDAR maps, capable of simultaneously performing pose estimation and metric localization;

Rank2Reward: Learning Shaped Reward Functions from Passive Video

Daniel Yang, Abhishek Gupta

Robotic IntelligenceReinforcement LearningVideo

🎯 What it does: Propose the Rank2Reward technique, which learns a reward function from raw videos without action information to guide robot behavior learning

Rapid Resistography with Passive Overhead-perching Mechanism in an Unmanned Aerial System for Wood Structure Inspection

S. Lee, S. Foong

Robotic Intelligence

🎯 What it does: Developed a drone platform equipped with a passive prism gripper and cable drill to enable rapid remote-controlled high-altitude nail drilling inspection of wooden structures;

RAPIDFlow: Recurrent Adaptable Pyramids with Iterative Decoding for Efficient Optical Flow Estimation

Henrique Morimitsu, Xu-Cheng Yin

Autonomous DrivingComputational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkOptical FlowImageVideoBenchmark

🎯 What it does: Proposed a model called RAPIDFlow aimed at efficiently estimating high-quality optical flow on embedded devices.

RASCAL: A Scalable, High-redundancy Robot for Automated Storage and Retrieval Systems

Richard Black, Hugh Williams

Robotic Intelligence

🎯 What it does: Designed a scalable, highly redundant automated storage and retrieval system robot (RASCAL) specifically for handling small-load items in structured environments, with its feasibility validated through the prototype implementation of a media storage robot in a data center.

RaSim: A Range-aware High-fidelity RGB-D Data Simulation Pipeline for Real-world Applications

Xingyu Liu, Xiangyang Ji

Data SynthesisImage

🎯 What it does: Developed RaSim, a range-aware high-fidelity RGB-D data simulation pipeline to generate realistic depth images and bridge the simulation-to-reality gap.

RaTrack: Moving Object Detection and Tracking with 4D Radar Point Cloud

Zhijun Pan, Chris Xiaoxuan Lu

Object DetectionObject TrackingPoint Cloud

🎯 What it does: Proposes an algorithm called RaTrack for moving object detection and tracking using 4D radar point clouds, focusing on motion segmentation and clustering, and equipped with a motion estimation module;

RB5 Low-Cost Explorer: Implementing Autonomous Long-Term Exploration on Low-Cost Robotic Hardware

Adam Seewald, A. Dollar

Robotic IntelligenceSimultaneous Localization and MappingImagePoint Cloud

🎯 What it does: Implemented a low-cost mobile robot RB5 equipped with an RGB-D camera and low-power computing hardware, utilizing a rocker-bogie suspension to perform long-term autonomous exploration in unknown, GPS-denied indoor and outdoor environments, and proposed an exploration strategy combining frontier and sampling methods, path-following vector fields, and advanced SLAM algorithms;

RBI-RRT*: Efficient Sampling-based Path Planning for High-dimensional State Space

Fang Chen, Sicong Liu

OptimizationRobotic Intelligence

🎯 What it does: Proposed a Reconstructed Bi-directional Informed RRT* (RBI-RRT*) algorithm for efficient sampling-based path planning in high-dimensional spaces.

RCM-Fusion: Radar-Camera Multi-Level Fusion for 3D Object Detection

Ji Song Kim, Junwon Choi

Object DetectionAutonomous DrivingMultimodalityPoint Cloud

🎯 What it does: Propose the RCM-Fusion radar-camera multi-level fusion model for 3D object detection, which includes feature-level fusion and instance-level fusion;

Real-Time Adaptive Safety-Critical Control with Gaussian Processes in High-Order Uncertain Models

Yu Zhang, A. Knoll

Optimization

🎯 What it does: Propose a real-time adaptive safety-critical control framework for systems with uncertain parameters, divided into two stages: the first stage constructs a sparse Gaussian process learning model with a forgetting factor and a single inducing point, while the second stage designs a safety filter based on higher-order control barrier functions and collaborates with the learning model to achieve real-time control under safety constraints.

Real-time Batched Distance Computation for Time-Optimal Safe Path Tracking

Shohei Fujii, Quang Pham

OptimizationComputational EfficiencyRobotic Intelligence

🎯 What it does: Propose a batch real-time distance calculation method based on precomputed link-local SDF, and use a neural network to accelerate preprocessing for time-optimal safe path tracking.

Real-Time Capable Decision Making for Autonomous Driving Using Reachable Sets

Niklas Kochdumper, Stanley Bak

Autonomous DrivingOptimizationBenchmark

🎯 What it does: This paper proposes a real-time decision module based on reachable sets for path selection and generation in autonomous driving.

Real-time Contact State Estimation in Shape Control of Deformable Linear Objects under Small Environmental Constraints

Kejia Chen, A. Knoll

Anomaly DetectionRobotic Intelligence

🎯 What it does: Proposed a real-time contact state estimation method for shape control of deformable linear objects under small environmental constraints.

Real-time Dexterous Prosthesis Hand Control by Decoding Neural Information Based on EMG Decomposition

Zhenzhi Ying, Naohiko Sugita

Robotic IntelligenceBiomedical Data

🎯 What it does: Proposed and implemented a neural information decoding method based on EMG decomposition for real-time control of dexterous movements of a prosthetic hand, and verified its feasibility through experiments.

Real-time Dynamic-consistent Motion Planning for Over-actuated UAVs

Yao Su, Hangxin Liu

Autonomous DrivingOptimizationComputational EfficiencyRobotic Intelligence

🎯 What it does: Proposed an efficient, real-time deployable over-actuated unmanned vehicle dynamic consistency motion planning method.

Real-Time Estimation for the Swimming Direction of Robotic Fish Based on IMU Sensors*

Shikun Li, Guangming Xie

Robotic IntelligenceTime Series

🎯 What it does: This study installs low-cost inertial measurement units (IMUs) on the head and tail of a two-jointed robotic fish, and proposes a method combining a Kalman filter to correct yaw drift and an anti-jitter estimation (ASE) algorithm to achieve real-time swimming direction estimation at a high frequency of 100 Hz. Subsequently, this method is applied to closed-loop control of swimming direction.

Real-Time Locomotion Transitions Detection: Maximizing Performances with Minimal Resources

Zeynep Ozge Orhan, Mohamed Bouri

OptimizationComputational Efficiency

🎯 What it does: Proposed a real-time motion transition detection method by training a machine learning model to set activity-specific thresholds for identifying transition moments between different gaits.

Real-Time Planning Under Uncertainty for AUVs Using Virtual Maps

Ivana Collado-Gonzalez, Brendan Englot

Autonomous DrivingOptimizationSimultaneous Localization and Mapping

🎯 What it does: Designed a real-time planning framework based on a virtual map for path planning of AUVs under uncertainty in sparse and large-scale underwater environments.

Real-time Whole-body Motion Planning for Mobile Manipulators Using Environment-adaptive Search and Spatial-temporal Optimization

Chengkai Wu, Boyu Zhou

OptimizationRobotic Intelligence

🎯 What it does: Proposed a whole-body motion planning method for mobile manipulators that can generate high-quality, safe, agile, and feasible trajectories in real time.

Real-to-Sim Deformable Object Manipulation: Optimizing Physics Models with Residual Mappings for Robotic Surgery

Xiao Liang, Michael C. Yip

Domain AdaptationOptimizationRobotic IntelligencePhysics Related

🎯 What it does: An online adaptive parameter tuning method is proposed, which estimates residual mapping to bridge the gap between reality and simulation, and online optimizes stiffness parameters to improve prediction accuracy in soft tissue manipulation.

Realistic Data Generation for 6D Pose Estimation of Surgical Instruments

Juan Antonio Barragan, P. Kazanzides

Data SynthesisPose EstimationImageBiomedical Data

🎯 What it does: Proposed an improved surgical robot simulation environment and automated data generation pipeline for generating 6D pose data, created a dataset of 7.5k surgical needle images with pose annotations, and evaluated a state-of-the-art pose estimation network.

Realtime Robust Shape Estimation of Deformable Linear Object

Jiaming Zhang, Mehran Armand

Pose Estimation

🎯 What it does: Real-time estimation of the shape of linearly deformable objects

Recasting Generic Pretrained Vision Transformers As Object-Centric Scene Encoders For Manipulation Policies

Jianing Qian, Dinesh Jayaraman

Representation LearningRobotic IntelligenceTransformerVision-Language-Action ModelImage

🎯 What it does: Proposes the SOFT (Scene Objects From Transformers) framework, which leverages the attention mechanism of a pre-trained visual Transformer (PVT) model to locate and describe objects, thereby generating object-centric embeddings for robot manipulation.

Receding-Constraint Model Predictive Control using a Learned Approximate Control-Invariant Set

Gianni Lunardi, Andrea Del Prete

Optimization

🎯 What it does: Proposed a recursive constraint MPC method based on learning-driven approximate control invariant sets, achieving recursive feasibility and safety guarantees.

Recency Bias in Task Performance History Affects Perceptions of Robot Competence and Trustworthiness

Matthew B. Luebbers, Bradley Hayes

Robotic IntelligenceVideo

🎯 What it does: Conducted an experiment with 53 participants to investigate how the sequence of robot tasks affects human recall bias (i.e., decreasing bias) regarding robot capabilities and trust

Reciprocal and Non-Reciprocal Swarmalators with Programmable Locomotion and Formations for Robot Swarms

Steven Ceron, Daniela Rus

Robotic Intelligence

🎯 What it does: Studied the reciprocity and non-reciprocity in adaptive interaction models (swarmalators), and achieved programmable motion and spatial distribution of robot swarms through non-reciprocal coupling and control barrier functions.

RecNet: An Invertible Point Cloud Encoding through Range Image Embeddings for Multi-Robot Map Sharing and Reconstruction

Nikolaos Stathoulopoulos, G. Nikolakopoulos

CompressionRobotic IntelligenceAuto EncoderPoint Cloud

🎯 What it does: Propose RecNet, which projects 3D point clouds into range images and uses an encoder-decoder framework for compression and reconstruction, generating reversible point cloud encoding and utilizing latent vectors for efficient pose recognition.

Reconfiguration of a 2D Structure Using Spatio-Temporal Planning and Load Transferring

Javier Garcia, Aaron T. Becker

OptimizationRobotic Intelligence

🎯 What it does: Studied the reconfiguration of two-dimensional structural materials by collaborative robot groups and proposed two methods.

Recursive Least Squares with Log-Determinant Divergence Regularisation for Online Inertia Identification

Namhoon Cho, Hyo-Sang Shin

OptimizationRobotic IntelligenceTime SeriesPhysics Related

🎯 What it does: Propose a recursive algorithm using regularized least squares to solve the online rigid body dynamics parameter identification problem, emphasizing the physical consistency of inertial parameters.

Reducing Non-IID Effects in Federated Autonomous Driving with Contrastive Divergence Loss

Tuong Khanh Long Do, A. Nguyen

Autonomous DrivingFederated LearningContrastive Learning

🎯 What it does: Proposed a new contrastive divergence loss to address the non-IID problem of autonomous driving data in federated learning environments

ReefGlider: A Highly Maneuverable Vectored Buoyancy Engine Based Underwater Robot

Kevin Macauley, Y. Girdhar

Robotic Intelligence

🎯 What it does: Proposes a new underwater robot named ReefGlider that uses only buoyancy control to achieve high maneuverability, aiming to bridge the capability gap between traditional propulsion-driven AUVs and conventional buoyancy-driven underwater gliders.

Refining Pre-Trained Motion Models

Xinglong Sun, Leonidas J. Guibas

Object TrackingSupervised Fine-TuningOptical Flow

🎯 What it does: Improved existing supervised motion estimation models by employing a self-supervised two-stage training process to fine-tune pre-trained models;

REFORMA: Robust REinFORceMent Learning via Adaptive Adversary for Drones Flying under Disturbances

Hao-Lun Hsu, Miroslav Pajic

Robotic IntelligenceReinforcement Learning

🎯 What it does: Proposes REFORMA, a reinforcement learning framework that utilizes adaptive adversaries for real-time online adaptation to achieve robust control of drones under unknown perturbations.

Reg-NF: Efficient Registration of Implicit Surfaces within Neural Fields

Stephen Hausler, P. Moghadam

Pose EstimationOptimizationComputational EfficiencyNeural Radiance Field

🎯 What it does: Developed Reg-NF for optimizing relative 6-DoF transformation registration between two arbitrary neural fields.

Region-determined localization method for unmanned ground vehicle under pole-like feature environment

Yu-Hsiang Lai, Feng-Li Lian

Autonomous DrivingSimultaneous Localization and Mapping

🎯 What it does: A region-determined localization method for unmanned ground vehicles is proposed, utilizing rod-like features (such as trees or streetlights) to address GNSS degradation localization issues, consisting of three modules: mapping, boundary determination, and localization.

Regrasping on Printed Circuit Boards with the Smart Suction Cup

Jungpyo Lee, Hannah S. Stuart

OptimizationRobotic IntelligenceImage

🎯 What it does: This paper studies improving the grasping performance of suction cups on printed circuit boards (PCB) through regrasping control, conducting experiments using a vision-based grasping planner under two settings: static and conveyor belt, targeting PCBs with large surface features for classification;

Regressing Transformers for Data-efficient Visual Place Recognition

María Leyva-Vallina, N. Petkov

RecognitionRetrievalTransformerImage

🎯 What it does: Propose to treat visual place recognition as a regression problem, using camera field-of-view overlap as the ground truth for similarity to train image descriptors.

Reinforcement Learning for Blind Stair Climbing with Legged and Wheeled-Legged Robots

Simon Chamorro, Roland Siegwart

Robotic IntelligenceReinforcement Learning

🎯 What it does: Propose a general-purpose staircase climbing controller for upright or wheeled legged robots based on reinforcement learning, employing position-based task formulation and achieving real-world deployment without external sensors through simulation training;

Reinforcement Learning for Collision-free Flight Exploiting Deep Collision Encoding

M. Kulkarni, Kostas Alexis

Robotic IntelligenceConvolutional Neural NetworkReinforcement LearningImage

🎯 What it does: A modular deep navigation strategy based on deep collision encoding and reinforcement learning was developed to achieve collision-free flight for drones.

Reinforcement learning for freeform robot design

Muhan Li, Sam Kriegman

Robotic IntelligenceReinforcement Learning

🎯 What it does: Using policy gradient methods, train robots to design arbitrary external and internal structures in free-form space, where actions include adding or removing atomic building blocks to form macroscopic structures.

Reinforcement Learning for Reduced-order Models of Legged Robots

Yu-Ming Chen, Michael Posa

OptimizationRobotic IntelligenceReinforcement Learning

🎯 What it does: Propose a model-based reinforcement learning framework on a bipedal platform, learning a reduced-order model (ROM) to embed into model predictive control (MPC).