ICRA 2023 Papers — Page 11
IEEE International Conference on Robotics and Automation · 1341 papers
Reinforcement Learning Control of a Reconfigurable Planar Cable Driven Parallel Manipulator
Adhiti Raman, V. Krovi
Robotic IntelligenceReinforcement Learning
🎯 What it does: The study employs deep reinforcement learning to achieve dynamic trajectory tracking control for a reconfigurable planar cable-driven parallel robot, and explores three different RL implementation setups (standard non-redundant CDPR, end-to-end redundant resolved reconfigurable CDPR, and a decoupling method separating dynamics and geometric redundancy).
Reinforcement Learning for Laser Welding Speed Control Minimizing Bead Width Error
Toshimitsu Kaneko, T. Sakai
OptimizationReinforcement Learning
🎯 What it does: Proposes a laser welding speed control method based on reinforcement learning, adopting a reward discount based on welding length to improve the optimization of seam width error.
Reinforcement Learning for Safe Robot Control using Control Lyapunov Barrier Functions
Desong Du, Wei Pan
Safty and PrivacyRobotic IntelligenceReinforcement Learning
🎯 What it does: Propose a safety and reachability analysis method based on control Lyapunov barrier functions, and develop a model-free Lyapunov Barrier Actor-Critic (LBAC) algorithm to search for controllers that satisfy these constraints.
Reinforcement Learning with Probabilistically Safe Control Barrier Functions for Ramp Merging
Soumith Udatha, J. Dolan
Autonomous DrivingReinforcement Learning
🎯 What it does: Embed probabilistic safety control barrier functions into reinforcement learning strategies, proposing a safety-assured policy optimization algorithm SAPO-RM for slope merging scenarios.
Reinforcement Learning-Based Optimal Multiple Waypoint Navigation
Christos Vlachos, K. Kyriakopoulos
OptimizationReinforcement Learning
🎯 What it does: Proposed a multi-endpoint optimal motion planning method based on artificial potential fields and reinforcement learning
Relay Pursuit for Multirobot Target Tracking on Tile Graphs
Shashwata Mandal, S. Bhattacharya
Object TrackingGraph
🎯 What it does: Proposed a target tracking strategy based on multi-robot vision, utilizing a tile map to implement a dynamic tracking scheme for multiple robots to track moving intruders in polygonal environments.
Rendezvous and Docking of Magnetic Helical Microrobots Along Arc Orbits for Field-directed Assembly and Disassembly
Shuideng Wang, Lixin Dong
Robotic IntelligencePhysics Related
🎯 What it does: Proposed a scheme to control magnetic helical microrobots to achieve aggregation and disassembly along an arc trajectory under a uniform rotating magnetic field.
Repetitive Twisting Durability of Synthetic Fiber Ropes
Shinya Sadachika, G. Endo
Robotic IntelligencePhysics Related
🎯 What it does: Experimental tests were conducted on synthetic fiber ropes made from five different fibers under repeated torsion conditions to evaluate their durability and compare the performance of single ropes and parallel double ropes made of Dyneema.
Resilient Terrain Navigation with a 5 DOF Metal Detector Drone
Patrick Pfreundschuh, Olov Andersson
OptimizationRobotic IntelligenceSimultaneous Localization and MappingPoint CloudSequential
🎯 What it does: Developed a metal detector autonomous inspection system based on a 5-DOF drone, capable of detecting metal targets on uneven and occluded terrain.
ResiPlan: Closing the Planning-Acting Loop for Safe Underwater Navigation
Marios Xanthidis, K. Alexis
Robotic Intelligence
🎯 What it does: Proposes the ResiPlan framework, achieving closed-loop planning and execution in varying underwater environments, thereby enhancing the safety navigation capability of autonomous submarines.
Reslicing Ultrasound Images for Data Augmentation and Vessel Reconstruction
Cecilia G. Morales, A. Dubrawski
SegmentationGenerationData SynthesisImageBiomedical DataUltrasound
🎯 What it does: Proposed and implemented RESUS (RESlicing of UltraSound Images), a weakly supervised data augmentation technique, which generates perspectives difficult to obtain in vivo by slicing from three-dimensional volumes reconstructed from tracked two-dimensional ultrasound images for training semantic segmentation models.
Resolution Complete In-Place Object Retrieval given Known Object Models
Daniel Nakhimovich, Kostas E. Bekris
OptimizationRobotic IntelligenceImagePoint CloudMesh
🎯 What it does: Proposed a robot task planning framework based on known object 3D models for safely retrieving target objects from stacked obstacles in confined workspaces using grasp and workspace placement actions;
Reuse your features: unifying retrieval and feature-metric alignment
Javier Morlana, J. Montiel
Pose EstimationRetrievalImage
🎯 What it does: Proposed a compact visual localization pipeline that unifies all steps including image retrieval, candidate re-ranking, initial pose estimation, and camera pose refinement, and implemented the DRAN (Deep Retrieval and Image Alignment Network) model;
RFFCE: Residual Feature Fusion and Confidence Evaluation Network for 6DoF Pose Estimation
Qiwei Meng, Wei Song
Pose EstimationConvolutional Neural NetworkMultimodality
🎯 What it does: Proposed a two-stage 6DoF pose estimation network called RFFCE based on RGBD, which uses a deep network to extract features and match object points, and then calculates the final pose using geometric principles.
RGB-D Grasp Detection via Depth Guided Learning with Cross-modal Attention
Ran Qin, Di Huang
Robotic IntelligenceConvolutional Neural NetworkImageMultimodality
🎯 What it does: Proposed a learning-based RGB-D grasping detection method called Depth Guided Cross-modal Attention Network (DGCAN), which employs a complete 6D rectangular representation and incorporates grasping depth prediction, while designing a Local Cross-modal Attention (LCA) module to correct depth features and fuse them with RGB features, and validated the effectiveness of the method in simulation and real-world evaluations.
RGB-Event Fusion for Moving Object Detection in Autonomous Driving
Zhuyun Zhou, D. Ginhac
Object DetectionAutonomous DrivingMultimodality
🎯 What it does: Proposes a RGB-Event fusion network called RENet to enhance the performance of moving object detection in autonomous driving environments.
RGB-Only Reconstruction of Tabletop Scenes for Collision-Free Manipulator Control
Z-H. Tang, Stan Birchfield
Robotic IntelligenceNeural Radiance FieldImage
🎯 What it does: Developed a collision-free control system for robotic manipulators using only RGB views.
Risk-Aware Model Predictive Path Integral Control Using Conditional Value-at-Risk
Ji Yin, P. Tsiotras
OptimizationRobotic Intelligence
🎯 What it does: A risk-aware model predictive control method based on Conditional Value-at-Risk (CVaR) is proposed for path integral control, addressing arbitrary uncertainties in robot planning and control.
Risk-Aware Neural Navigation From BEV Input for Interactive Driving
Suzanna Jiwani, Daniela Rus
Autonomous DrivingReinforcement LearningImage
🎯 What it does: Construct an interpretable risk representation based on the bird's-eye view (BEV) and generate risk-averse trajectories, including a risk map generator, differentiable value iteration strategy learning, and a trajectory sampler;
Risk-aware Path Planning via Probabilistic Fusion of Traversability Prediction for Planetary Rovers on Heterogeneous Terrains
Masafumi Endo, G. Ishigami
Robotic Intelligence
🎯 What it does: Propose a new path planning algorithm that explicitly considers machine learning prediction errors. It obtains a multi-modal slip distribution by probabilistically fusing terrain type classification with slip prediction models, performs risk assessment based on this distribution, and generates risk-aware travel costs.
Risk-aware Recharging Rendezvous for a Collaborative Team of UAVs and UGVs
A. Asghar, Pratap Tokekar
OptimizationRobotic Intelligence
🎯 What it does: Proposed and studied the risk-aware charging docking problem for a collaborative team of unmanned aerial vehicles (UAV) and unmanned ground vehicles (UGV), and presented a bi-objective approximation algorithm
Risk-aware Spatio-temporal Logic Planning in Gaussian Belief Spaces
M. Vahs, Jana Tumova
Robotic Intelligence
🎯 What it does: A risk-aware spatiotemporal logic planning method was designed and implemented in a Gaussian belief space, minimizing the risk of regulation violations during task execution and enabling robots to actively collect state information.
RLAfford: End-to-End Affordance Learning for Robotic Manipulation
Yiran Geng, Hao Dong
Robotic IntelligenceReinforcement Learning
🎯 What it does: Explored using contact information generated during RL training as a unified visual representation to predict contact maps, achieving an end-to-end affordance learning framework that generalizes across various robotic manipulation tasks.
Rmagine: 3D Range Sensor Simulation in Polygonal Maps via Ray Tracing for Embedded Hardware on Mobile Robots
Alexander Mock, J. Hertzberg
Computational EfficiencyRobotic IntelligencePoint CloudMesh
🎯 What it does: Proposes a library called Rmagine that directly uses ray tracing to simulate range sensors on triangular mesh environment maps on mobile robot embedded hardware.
Road Anomaly Segmentation Based on Pixel-wise Logit Variance with Iterative Background Highlighting
Dongkun Lee, Ho-Jin Choi
SegmentationAnomaly DetectionAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: This study proposes an anomaly segmentation method based on pixel-level logit variance and iterative background highlighting, utilizing logit information from pre-trained semantic segmentation networks to identify abnormal regions in urban scenes.
RoboSC: a domain-specific language for supervisory controller synthesis of ROS applications
B. Wesselink, E. Torta
Robotic Intelligence
🎯 What it does: Proposes a domain-specific language (DSL) called RoboSC for the synthesis of supervisory controllers in ROS applications, supporting ROS/ROS2 and supervisory control theory, and automatically generating all artifacts required for integration and deployment.
Robot explanatory narratives of collaborative and adaptive experiences
Alberto Olivares-Alarcos, G. Alenyà
Explainability and InterpretabilityRobotic IntelligenceLarge Language ModelText
🎯 What it does: Proposes a narrative method integrating collaborative robot ontologies, event memory, and knowledge extraction algorithms for storing, retrieving, and generating textual narratives describing robot collaboration and adaptation processes.
Robot Mimicry Attack on Keystroke-Dynamics User Identification and Authentication System
Rongyu Yu, M. Imran
Adversarial AttackRobotic IntelligenceReinforcement LearningTextAudio
🎯 What it does: Studied the impact of robot-simulated attacks on keyboard dynamic user identification and authentication systems, proposed and implemented an attack framework based on deep Q-networks, and evaluated it on a real robot testbed.
Robot Person Following Under Partial Occlusion
Hanjing Ye, Hong Zhang
Object TrackingPose EstimationRobotic IntelligenceVideo
🎯 What it does: Propose a localization method based on visible joints from a monocular camera, enabling robot tracking of humans under partial occlusion
Robot Trust and Self-Confidence Based Role Arbitration Method for Physical Human-Robot Collaboration
Qiao Wang, Chin-Teng Lin
Robotic Intelligence
🎯 What it does: Investigated a trust-based role arbitration method, proposed a computational model of robot trust and confidence (TSC), and validated the effectiveness of this method in physical human-robot collaboration through human-robot loop experiments.
Robot-Assisted Eye-Hand Coordination Training System by Estimating Motion Direction Using Smooth-Pursuit Eye Movements
Xiao Li, Ai-Guon Song
Robotic Intelligence
🎯 What it does: Developed a robot-assisted eye-hand coordination training system based on smooth pursuit eye movement estimation of motion direction
Robotic Control Using Model Based Meta Adaption
Karam Daaboul, J. M. Zöllner
Robotic IntelligenceMeta LearningReinforcement LearningWorld Model
🎯 What it does: Proposed the Meta Adaptation Controller (MAC), which utilizes Meta Reinforcement Learning (MRL) to transfer preferred robotic behaviors learned from previous tasks to multiple similar tasks, and rapidly adapts to dynamic changes by identifying actions that achieve similar outcomes in new tasks.
Robotic Fastening with a Manual Screwdriver
Ling Tang, Yan-Bin Jia
Robotic Intelligence
🎯 What it does: This paper studies how a robotic arm assembles a fixed-configuration screwdriver into a screw pre-installed in a threaded hole, and uses the tool to tighten the screw.
Robotic Method and Instrument to Efficiently Synthesize Faulty Conditions and Mass-Produce Faulty-Conditioned Data for Rotary Machines
Yip Fun Yeung, K. Toumi
Data SynthesisRobotic Intelligence
🎯 What it does: Developed a fault condition synthesis method based on robot force control and a specialized manipulator, achieving rapid synthesis and large-scale data generation for rotating machine fault conditions.
Robotic Navigation Autonomy for Subretinal Injection via Intelligent Real-Time Virtual iOCT Volume Slicing
Shervin Dehghani, I. Iordachita
Pose EstimationRobotic IntelligenceConvolutional Neural NetworkBiomedical Data
🎯 What it does: Proposed an autonomous robotic navigation framework based on real-time iOCT volumetric slicing for precise subretinal injection operations.
Robotic Sonographer: Autonomous Robotic Ultrasound using Domain Expertise in Bayesian Optimization
Deepak Raina, S. Saha
Robotic IntelligenceConvolutional Neural NetworkBiomedical DataUltrasound
🎯 What it does: This paper proposes an autonomous robot ultrasound system based on Bayesian optimization and domain expertise for predicting and scanning regions that can yield diagnostic quality ultrasound images.
Robotic Table Wiping via Reinforcement Learning and Whole-body Trajectory Optimization
T. Lew, Montserrat González
OptimizationRobotic IntelligenceReinforcement LearningImageStochastic Differential Equation
🎯 What it does: Proposed a framework enabling multi-purpose assistive mobile robots to autonomously wipe tables and clean spills and debris;
RobotSweater: Scalable, Generalizable, and Customizable Machine-Knitted Tactile Skins for Robots
Zilin Si, Wenzhen Yuan
Robotic Intelligence
🎯 What it does: Developed a machine-knitted pressure-sensitive tactile skin called RobotSweater, which can easily adhere to robot surfaces and was validated for contact detection, multi-point localization, and pressure sensing functions on both flat and curved test platforms; subsequently, the tactile skin was used to achieve closed-loop control in two scenarios (human-guided robotic arm control and human-robot interaction in mobile robots).
Robust Bipedal Locomotion: Leveraging Saltation Matrices for Gait Optimization
Maegan Tucker, A. Ames
OptimizationRobotic Intelligence
🎯 What it does: By incorporating the saltation matrix into the Hybrid Zero Dynamics framework and jointly minimizing the saltation matrix norm and robot torque, a more robust gait is generated.
Robust co-design of robots via cascaded optimisation
Akhil Sathuluri, M. Zimmermann
OptimizationRobotic Intelligence
🎯 What it does: Identify non-intuitive designs using a two-step cascading optimization method and construct a solution space to recover performance loss
Robust Collaborative 3D Object Detection in Presence of Pose Errors
Yifan Lu, Yanfeng Wang
Object DetectionGraph Neural NetworkPoint Cloud
🎯 What it does: Propose the CoAlign framework, which utilizes proxy-target pose graph modeling and multi-scale feature fusion to enhance collaborative 3D object detection performance under pose errors.
Robust Double-Encoder Network for RGB-D Panoptic Segmentation
Matteo Sodano, C. Stachniss
SegmentationConvolutional Neural NetworkMultimodality
🎯 What it does: Proposed a dual encoder network that separately encodes RGB and depth, and fuses features at different resolutions;
Robust Forecasting for Robotic Control: A Game-Theoretic Approach
Shubhankar Agarwal, Sandeep P. Chinchali
Robotic IntelligenceTime Series
🎯 What it does: Proposed a zero-sum two-player game framework where the robot predictor competes against an imaginary adversary to generate robust time series predictions, addressing the impact of noise, outliers, and out-of-sample scenarios.
Robust Human Pose Estimation under Gaussian Noise
Patrick Schlosser, C. Ledermann
Pose EstimationConvolutional Neural NetworkImage
🎯 What it does: Evaluated the impact of Gaussian noise on human pose estimation and proposed two countermeasures: denoising images and robust training.
Robust Imaging Sonar-based Place Recognition and Localization in Underwater Environments
Hogyun Kim, Younggun Cho
Pose EstimationSimultaneous Localization and MappingImage
🎯 What it does: A robust pose recognition and loop closure method based on imaging SONAR is proposed, achieving pose estimation and loop closure correction through geometric information encoding of raw SONAR measurements, hierarchical search, adaptive translation and filling, and ICP (Iterative Closest Point).
Robust Incremental Smoothing and Mapping (riSAM)
D. McGann, M. Kaess
OptimizationSimultaneous Localization and Mapping
🎯 What it does: Proposes a robust incremental bundle adjustment and mapping algorithm (riSAM) based on Graduated Non-Convexity for robust optimization in online incremental SLAM.
Robust Locomotion on Legged Robots through Planning on Motion Primitive Graphs
Wyatt Ubellacker, A. Ames
Robotic Intelligence
🎯 What it does: Propose an online planning method based on motion primitive graphs to achieve robust locomotion for quadruped robots in disturbed environments.
Robust MADER: Decentralized and Asynchronous Multiagent Trajectory Planner Robust to Communication Delay
Kota Kondo, J. How
OptimizationRobotic Intelligence
🎯 What it does: Propose a decentralized, asynchronous multi-agent trajectory planner called RMADER that can handle communication delays.
Robust Map Fusion with Visual Attention Utilizing Multi-agent Rendezvous
Jaein Kim, Byoung-Tak Zhang
Pose EstimationRobotic IntelligenceTransformerSimultaneous Localization and MappingImage
🎯 What it does: Propose a robust map fusion system based on visual attention that can effectively fuse local maps in multi-robot collaborative localization and mapping (SLAM), even without shared information.
Robust Navigation with Cross-Modal Fusion and Knowledge Transfer
Wenzhe Cai, Changyin Sun
Knowledge DistillationRobotic IntelligenceSimultaneous Localization and MappingMultimodality
🎯 What it does: Achieving robustness and generality in mobile robot navigation through cross-modal fusion and knowledge transfer
Robust Output Feedback controller for a Serial Robotic Manipulator with Unknown Nonlinearities and External Disturbances
M. Al Saaideh, M. Al Janaideh
Robotic Intelligence
🎯 What it does: A robust output feedback controller is proposed for multi-link serial manipulators with unknown nonlinear dynamics and external disturbances, enabling the end-effector to track the desired trajectory under unknown system dynamics and disturbances.
Robust Plant Localization and Phenotyping in Dense 3D Point Clouds for Precision Agriculture
H. J. Nelson, N. Papanikolopoulos
Object DetectionSegmentationPoint CloudAgriculture Related
🎯 What it does: Proposes a multi-stage unsupervised method that automatically detects the individual positions and heights of field crops using dense 3D point clouds generated by drone's regular cameras, enabling automated analysis of crop growth stages.
Robust Robot Planning for Human-Robot Collaboration
Yang You, O. Buffet
Robotic Intelligence
🎯 What it does: Propose a method for automatically generating uncertain human behavior strategies while considering robot behavior, and present a robust robot planning algorithm based on distributed human behavior using POMDP
Robust Uncertainty Estimation for Classification of Maritime Objects
J. Becktor, L. Nalpantidis
ClassificationAnomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: Explore uncertainty estimation in the maritime domain, proposing a unified method that combines intra-class uncertainty obtained via Monte Carlo Dropout with the latest anomaly detection techniques, validated on CIFAR10 and a self-built SHIPS dataset.
Robust, High-Rate Trajectory Tracking on Insect-Scale Soft-Actuated Aerial Robots with Deep-Learned Tube MPC
Andrea Tagliabue, Department of Mechanical Engineering
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: Proposed an agile and computationally efficient trajectory tracking scheme for sub-clone level soft-actuated drones, employing a cascaded control structure comprising an adaptive attitude controller and a neural network strategy;
RoLM: Radar on LiDAR Map Localization
Yukai Ma, Yong Liu
Autonomous DrivingSimultaneous Localization and MappingMultimodalityPoint Cloud
🎯 What it does: Propose a heterogeneous localization method (RoLM) that aligns radar with LiDAR maps, achieving real-time elimination of radar odometry drift, improving localization accuracy without requiring loop closure.
ROSMC: A High-Level Mission Operation Framework for Heterogeneous Robotic Teams
Ryo Sakagami, F. Stulp
Robotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposed the ROS-MC high-level task operation framework for task synchronization and operation in heterogeneous mobile robot teams.
RoSS: Rotation-induced Aliasing for Audio Source Separation
Hyungjoo Seo, Romit Roy Choudhury
Audio
🎯 What it does: Proposed and implemented the RoSS method for audio source separation utilizing microphone array rotation to generate delay aliasing, and validated its effectiveness on a complete prototype.
Rotation Synchronization via Deep Matrix Factorization
Gk Tejus, F. Arrigoni
Pose EstimationOptimization
🎯 What it does: Solved the rotation synchronization problem based on paired rotations, proposing an unsupervised neural network framework based on deep matrix decomposition
RPGD: A Small-Batch Parallel Gradient Descent Optimizer with Explorative Resampling for Nonlinear Model Predictive Control
Frederik Heetmeyer, Tobi Delbruck
OptimizationBenchmark
🎯 What it does: Proposed RPGD (Resampling Parallel Gradient Descent) optimizer for non-convex optimization in nonlinear model predictive control.
RTAW: An Attention Inspired Reinforcement Learning Method for Multi-Robot Task Allocation in Warehouse Environments
Aakriti Agrawal, Dinesh Manocha
Robotic IntelligenceTransformerReinforcement Learning
🎯 What it does: Study the multi-robot warehouse task allocation problem and propose the RTAW algorithm based on reinforcement learning
S*: On Safe and Time Efficient Robot Motion Planning
Riddhiman Laha, Sami Haddadin
OptimizationRobotic Intelligence
🎯 What it does: Proposed a safety and time-efficient robot motion planning algorithm S* based on graph search, which introduces an information-based cost balancing criterion to balance the shortest time path and higher safety speed path, thereby minimizing the overall planning time.
SACPlanner: Real-World Collision Avoidance with a Soft Actor Critic Local Planner and Polar State Representations
Khaled Nakhleh, Karina Palyutina
Robotic IntelligenceReinforcement Learning
🎯 What it does: Studied the performance of a ROS local planner based on reinforcement learning during the training process and its trajectory generation on real robots.
Safe and Distributed Multi-Agent Motion Planning under Minimum Speed Constraints
Inkyu Jang, H. J. Kim
Autonomous DrivingOptimizationSafty and Privacy
🎯 What it does: Proposed a distributed multi-agent motion planner to ensure collision avoidance and continuous feasibility for non-halt agents in obstacle-dense environments.
Safe and Efficient Navigation in Extreme Environments using Semantic Belief Graphs
M. Ginting, Ali-akbar Agha-mohammadi
Robotic IntelligenceGraph Neural Network
🎯 What it does: Propose a semantics-based planning method under uncertain perception environments, achieving safe and efficient navigation for robots in unknown, unstructured environments through the construction of a Semantic Belief Graph.
Safe Bipedal Path Planning via Control Barrier Functions for Polynomial Shape Obstacles Estimated Using Logistic Regression
Chengyang Peng, Ayonga Hereid
OptimizationSafty and PrivacyRobotic Intelligence
🎯 What it does: Proposes an RRT* algorithm based on Control Barrier Functions (CBF) for biped robots to generate safe and dynamically feasible paths in polynomial-shaped obstacle environments, and enhances free space exploration efficiency through a multi-step CBF navigation controller.
Safe Control using Vision-based Control Barrier Function (V-CBF)
Hossein Abdi, R. Ghabcheloo
Image TranslationAutonomous DrivingImage
🎯 What it does: Proposed a vision-based control barrier function (V-CBF) to achieve safe motion control in unknown environments.
Safe Model-based Control from Signal Temporal Logic Specifications Using Recurrent Neural Networks
Wenliang Liu, C. Belta
Recurrent Neural Network
🎯 What it does: Propose a policy search method for learning a controller from Signal Temporal Logic (STL) specifications, jointly learning an unknown affine control system model and control policy.
Safe Operations of an Aerial Swarm via a Cobot Human Swarm Interface
Sydrak S. Abdi, D. Paley
Safty and PrivacyRobotic Intelligence
🎯 What it does: Propose an interactive and control method for quadrotor drone swarms within confined spaces using EMG gesture control, employing a velocity controller based on distance potential functions to prevent collisions, and providing state feedback to the operator via a vibratory tactile vest.
Safe Real-World Autonomous Driving by Learning to Predict and Plan with a Mixture of Experts
S. Pini, Sergey Zagoruyko
Autonomous DrivingMixture of ExpertsSequential
🎯 What it does: This paper proposes a unified neural network architecture that models multiple future trajectory distributions for both the vehicle itself and other road agents, and selects the trajectory with the lowest safety cost and highest probability during inference.
Safe Reinforcement Learning of Dynamic High-Dimensional Robotic Tasks: Navigation, Manipulation, Interaction
Puze Liu, G. Chalvatzaki
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a new framework for safe exploration on the tangent space of a constraint manifold, which transforms the action space using the tangent space to ensure local satisfaction of safety constraints, and applied it to various robot platforms, including dynamic human models; achieved state-of-the-art performance in simulating high-dimensional dynamic tasks, and realized safe real-world deployment on the TIAGo++ robot, completing manipulation and human-robot interaction tasks.
Safe Self-Supervised Learning in Real of Visuo-Tactile Feedback Policies for Industrial Insertion
Letian Fu, Sachin Chitta
Robotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningMultimodality
🎯 What it does: Propose a vision-tactile feedback insertion strategy for safe self-supervised learning in real environments, capable of handling variations in grasp posture while reducing human input and collisions
Safeguarding Learning-Based Planners Under Motion and Sensing Uncertainties Using Reachability Analysis
Akshay Shetty, G. Gao
Safty and PrivacyRobotic Intelligence
🎯 What it does: Use reachability analysis to predict trajectory deviations under motion and perception uncertainties, and assess safety along the reference trajectory; subsequently, apply the analysis results to protect the learning-based trajectory planner, enhancing its robustness in real-world environments.
Safety Evaluation of Robot Systems via Uncertainty Quantification
W.-J. Baek, T. Kröger
Safty and PrivacyRobotic Intelligence
🎯 What it does: Proposes an online data-driven uncertainty quantification method to assess the safety of robotic systems in collaborative environments.
Safety Under Uncertainty: Tight Bounds with Risk-Aware Control Barrier Functions
Mitchell Black, Dimitra Panagou
Autonomous Driving
🎯 What it does: Propose a risk-aware control barrier function (RA-CBF) and prove that it provides a tighter upper bound on the system's unsafe probability within finite time.
Safety-Aware Unsupervised Skill Discovery
Sunin Kim, Julien Perez
Reinforcement Learning
🎯 What it does: Proposed an unsupervised skill discovery algorithm that learns a reusable and inherently safe skill library, and utilizes this safe skill library for hierarchical reinforcement learning to solve downstream tasks.
Safety-Constrained Policy Transfer with Successor Features
Zeyu Feng, Harold Soh
OptimizationReinforcement Learning
🎯 What it does: Proposed a Constrained Markov Decision Process (CMDP) framework to achieve safe policy transfer in reinforcement learning, separating task objectives through safety constraints and allowing explicit specification of multiple constraints.
Safety-Critical Controller Verification via Sim2Real Gap Quantification
Prithvi Akella, A. Ames
Robotic IntelligenceWorld Model
🎯 What it does: In the paper, the authors developed a program to quantify the sim2real error between models and real systems. Using this error, they constructed an uncertain model and synthesized and verified controllers in simulation using probabilistic verification methods, generating controllers with adjustable high safety probabilities on hardware. The process was validated on the Robotarium and quadruped robot platforms.
Safety-Critical Ergodic Exploration in Cluttered Environments via Control Barrier Functions
C. Lerch, Ian Abraham
OptimizationRobotic Intelligence
🎯 What it does: Proposed a safety-critical search trajectory planning method in crowded environments by combining discrete control barrier functions with trajectory optimization for coverage
SAMLoc: Structure-Aware Constraints With Multi-Task Distillation for Long-Term Visual Localization
Jian Ning, D. Kerr
Pose EstimationKnowledge DistillationRepresentation LearningSimultaneous Localization and MappingImage
🎯 What it does: Proposed a structure-aware and self-supervised visual localization system called SAMLoc to achieve fast and robust 6-DoF localization.
Sample Efficient Dynamics Learning for Symmetrical Legged Robots: Leveraging Physics Invariance and Geometric Symmetries
Jee-eun Lee, L. Sentis
Robotic Intelligence
🎯 What it does: Learn the dynamics of symmetric legged robots, propose a new dynamics learning method leveraging physical invariances and geometric symmetries, and achieve control of climbing robots based on learned inverse dynamics models
Sample-Driven Connectivity Learning for Motion Planning in Narrow Passages
Sihui Li, Neil T. Dantam
Robotic IntelligenceGraph
🎯 What it does: Proposes a learning-based sampling strategy to effectively generate sampling points in narrow passages of the configuration space, thereby improving overall planning time; the algorithm first learns connectivity information from the planning graph, then uses the learned results to generate sampling in narrow passages.
Sample-Efficient Goal-Conditioned Reinforcement Learning via Predictive Information Bottleneck for Goal Representation Learning
Qiming Zou, Einoshin Suzuki
Representation LearningReinforcement Learning
🎯 What it does: Proposed a self-supervised predictive information bottleneck method called PI-Goal for achieving sample-efficient goal-conditioned reinforcement learning in unknown goal spaces.
Sample, Crop, Track: Self-Supervised Mobile 3D Object Detection for Urban Driving LiDAR
Sangyun Shin, Niki Trigoni
Object DetectionAutonomous DrivingPoint CloudBenchmark
🎯 What it does: Proposed a self-supervised mobile 3D object detection method called SCT, which leverages motion cues and expected target size to predict dense 3D oriented bounding boxes in urban driving LiDAR point clouds, enhancing object discovery capabilities.
Sampling-based path planning under temporal logic constraints with real-time adaptation
Yizhou Chen, Ben M. Chen
OptimizationRobotic Intelligence
🎯 What it does: Proposes a sampling-based path planner capable of computing path schemes for temporal logic objectives and instantly adapting to non-static, partially unknown environments during robot execution.
Satellite Image Based Cross-view Localization for Autonomous Vehicle
Shan Wang, Hongdong Li
Autonomous DrivingContrastive LearningSimultaneous Localization and MappingImage
🎯 What it does: Using high-resolution satellite images as maps for cross-perspective vehicle localization.
Scalable Task-Driven Robotic Swarm Control via Collision Avoidance and Learning Mean-Field Control
Kai Cui, H. Koeppl
Robotic IntelligenceReinforcement Learning
🎯 What it does: Design a scalable task-based multi-robot control framework using mean field control and collision avoidance techniques, converting multi-agent control into single-agent distribution control and achieving collision-free group behavior.
SCAN: Socially-Aware Navigation Using Monte Carlo Tree Search
Jeongwoo Oh, Songhwai Oh
OptimizationRobotic Intelligence
🎯 What it does: Proposed the SCAN global planner, integrating human-robot interaction and future state prediction. It simulates future states and evaluates local goal quality through MCTS, uses Y-net for pedestrian motion prediction, and implements parallel simulation to accelerate computation.
SCARP: 3D Shape Completion in ARbitrary Poses for Improved Grasping
Bipasha Sen, M. Krishna
RestorationPose EstimationRobotic IntelligencePoint Cloud
🎯 What it does: Achieve 3D shape recovery under arbitrary poses by completing partial point clouds into full shapes and realizing pose estimation and normalization.
Scene-level Point Cloud Colorization with Semantics-and-geometry-aware Networks
Rongrong Gao, Qifeng Chen
GenerationPoint CloudBenchmark
🎯 What it does: Proposed a scene-level point cloud coloring network called SGNet based on semantic and geometric information, which can generate vivid and realistic colors for colorless three-dimensional point clouds;
SceneCalib: Automatic Targetless Calibration of Cameras and Lidars in Autonomous Driving
Ayon Sen, Ashraful Islam
Autonomous DrivingImagePoint Cloud
🎯 What it does: Propose SceneCalib, achieving automatic and target-free self-calibration of multi-camera and LiDAR systems.
SCORE: A Second-Order Conic Initialization for Range-Aided SLAM
Alan Papalia, J. Leonard
OptimizationSimultaneous Localization and Mapping
🎯 What it does: Proposes a novel initialization technique for distance-based SLAM (RA-SLAM), which first solves a second-order cone programming (SOCP) using convex relaxation methods, and then uses the results as the initialization for the original non-convex optimization.
SDF-Based Graph Convolutional Q-Networks for Rearrangement of Multiple Objects
Hogun Kee, Songhwai Oh
Robotic IntelligenceGraph Neural NetworkReinforcement LearningImageGraph
🎯 What it does: Propose a deep Q-learning framework based on symbolic distance fields (SDF) for non-grasping (e.g., pushing) rearrangement of multiple objects;
SE(3)-DiffusionFields: Learning smooth cost functions for joint grasp and motion optimization through diffusion
Julen Urain, G. Chalvatzaki
OptimizationRobotic IntelligenceDiffusion modelScore-based Model
🎯 What it does: Proposed a framework based on diffusion models to learn an SE(3) cost function for joint optimization in robotic grasping and motion planning;
Search Algorithms for Multi-Agent Teamwise Cooperative Path Finding
Z. Ren, H. Choset
Optimization
🎯 What it does: Proposes two algorithms, TC-CBS and TC-M*, for multi-team multi-objective path planning, implemented through extensions of the traditional MA-PF algorithms CBS and M*.
Security-Aware Reinforcement Learning under Linear Temporal Logic Specifications
Bohan Cui, Xiang Yin
Reinforcement Learning
🎯 What it does: The study addresses the reinforcement learning problem in Markov decision processes (MDPs) that satisfy linear temporal logic (LTL) specifications under safety constraints, and proposes a reward shaping method based on an initial state estimator and a limit-deterministic B"{u}chi automaton to achieve a control strategy that satisfies both the LTL task and ensures safety.
SEER: Safe Efficient Exploration for Aerial Robots using Learning to Predict Information Gain
Yuezhan Tao, Vijay R. Kumar
Robotic Intelligence
🎯 What it does: Designed a safe and efficient exploration framework that utilizes learning to predict information gain, aiding micro-drones in 3D exploration within indoor environments.
Segregator: Global Point Cloud Registration with Semantic and Geometric Cues
Peng Yin, Lihua Xie
Pose EstimationPoint Cloud
🎯 What it does: Proposed a global point cloud registration framework called Segregator, which efficiently constructs robust correspondences against anomalies and identifies inliers by leveraging semantic information and geometric distribution.
SEIL: Simulation-augmented Equivariant Imitation Learning
Ming Jia, Robert W. Platt
Data SynthesisReinforcement Learning
🎯 What it does: Proposed a method that integrates simulated transitions into expert trajectories and leverages O(2) symmetry for simulation-augmented equivariant imitation learning.
Self-Adaptive Driving in Nonstationary Environments through Conjectural Online Lookahead Adaptation
Tao Li, Quanyan Zhu
Autonomous DrivingRepresentation LearningMeta LearningReinforcement Learning
🎯 What it does: Proposes an online meta-reinforcement learning algorithm based on hypothesis-driven online prospective adaptation (COLA) for achieving adaptive driving in non-stationary environments.