ICRA 2023 Papers — Page 14
IEEE International Conference on Robotics and Automation · 1341 papers
V2XP-ASG: Generating Adversarial Scenes for Vehicle-to-Everything Perception
Hao Xiang, Jiaqi Ma
Autonomous DrivingAdversarial AttackGraph Neural NetworkPoint Cloud
🎯 What it does: Proposed an open-source adversarial scenario generator named V2XP-ASG for generating realistic and challenging traffic scenarios for LiDAR-based multi-agent perception systems.
Variable Admittance Interaction Control of UAVs via Deep Reinforcement Learning
Yuting Feng, Yixu Song
Robotic IntelligenceReinforcement Learning
🎯 What it does: Propose a reinforcement learning-based admittance control model, enabling drones to interact more effectively with external forces and automatically execute control tasks in unknown environments.
Vector Field Aided Trajectory Tracking by a 10-gram Flapping-Wing Micro Aerial Vehicle
Abdoullah Ndoye, F. Ruffier
Autonomous DrivingRobotic Intelligence
🎯 What it does: Implementing automatic trajectory tracking on a 10-gram flapping-wing micro aerial vehicle using a vector field method
Versatile Real-Time Motion Synthesis via Kino-Dynamic MPC With Hybrid-Systems DDP
He Li, Patrick M. Wensing
Robotic Intelligence
🎯 What it does: Propose a nonlinear model predictive control (NMPC) method that combines a hybrid dynamic model and a constrained differential dynamic programming (DDP) solver, enabling real-time replanning of specialized movements (e.g., jumping, obstacle crossing) and conventional gaits for quadruped robots, with its effectiveness verified on two platforms.
Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions
Chenhao Li, G. Martius
Robotic IntelligenceGenerative Adversarial Network
🎯 What it does: Proposed a cooperative adversarial method that employs self-supervised adversarial imitation learning to acquire a controllable multi-skill single versatile policy from an unlabeled mixed action dataset.
Video Waterdrop Removal via Spatio-Temporal Fusion in Driving Scenes
Q. Wen, Qifeng Chen
RestorationAutonomous DrivingVideo
🎯 What it does: Proposes an attention-based spatiotemporal fusion framework that uses multi-frame features to recover views occluded by water droplets.
Viewer-Centred Surface Completion for Unsupervised Domain Adaptation in 3D Object Detection
Darren Tsai, Stewart Worrall
Object DetectionDomain AdaptationPoint Cloud
🎯 What it does: Propose a view-based surface completion network (VCN) that unifies 3D detection object representations across different datasets under an unsupervised domain adaptation framework, to reduce performance degradation caused by differences in LiDAR scanning modes.
VINet: Visual and Inertial-based Terrain Classification and Adaptive Navigation over Unknown Terrain
Tianrui Guan, Liangjun Zhang
ClassificationRobotic IntelligenceSimultaneous Localization and MappingMultimodality
🎯 What it does: Propose a terrain classification network VINet based on visual and inertial information for robot navigation on different traversable surfaces
ViNL: Visual Navigation and Locomotion Over Obstacles
Simar Kareer, Joanne Truong
Robotic IntelligenceReinforcement LearningImage
🎯 What it does: Developed the ViNL system, integrating visual navigation and visual maneuvering strategies, enabling quadruped robots to navigate and cross small obstacles in unknown indoor environments using a front-facing camera
ViPFormer: Efficient Vision-and-Pointcloud Transformer for Unsupervised Pointcloud Understanding
Hongyu Sun, Deying Li
ClassificationSegmentationTransformerContrastive LearningImageMultimodalityPoint Cloud
🎯 What it does: Proposed a lightweight Vision-and-Pointcloud Transformer (ViPFormer) for unsupervised point cloud understanding, and pre-trained it to migrate to downstream tasks such as 3D shape classification and semantic segmentation.
Vis2Hap: Vision-based Haptic Rendering by Cross-modal Generation
Guanqun Cao, Shan Luo
GenerationData SynthesisImage
🎯 What it does: Proposes a vision-based tactile rendering method called Vis2Hap, which generates perceptible tactile sensations using visual inputs;
Visibility-Aware Navigation Among Movable Obstacles
Jose Muguira-Iturralde, Tomas Lozano-Perez
Autonomous DrivingOptimizationImage
🎯 What it does: Proposes the visibility-aware mobile obstacle navigation (VANAMO) problem, formalizes its definition, and introduces the LAMB algorithm to address it.
Vision-aided UAV Navigation and Dynamic Obstacle Avoidance using Gradient-based B-spline Trajectory Optimization
Zhefan Xu, K. Shimada
Autonomous DrivingOptimizationImagePoint Cloud
🎯 What it does: Propose a gradient B-spline trajectory optimization algorithm leveraging deep vision to achieve safe navigation and dynamic obstacle avoidance for drones in dynamic environments.
Vision-based Six-Dimensional Peg-in-Hole for Practical Connector Insertion
Kun Zhang, Wei Zhang
Pose EstimationRobotic IntelligenceImage
🎯 What it does: Study the 6D perception-based insertion problem, develop a robot assembly system based on a handheld RGB-D camera, combine learning detection with model-based pose estimation, model the insertion hole using rectangular features, and complete the insertion task of RJ45/HDMI slots on the KUKA iiwa7 robot.
Visual Affordance Prediction for Guiding Robot Exploration
Homanga Bharadhwaj, Shubham Tulsiani
Robotic IntelligenceTransformerAuto Encoder
🎯 What it does: Learning visual affordances and predicting the distribution of feasible future states generated by interacting with the scene
Visual Backtracking Teleoperation: A Data Collection Protocol for Offline Image-Based Reinforcement Learning
David Brandfonbrener, Jacob Varley
Data-Centric LearningRobotic IntelligenceReinforcement LearningImage
🎯 What it does: Developed the Visual Backtracking Teleoperation (VBT) protocol, which collects visually similar failure, recovery, and success samples, and uses these data to train image-based value functions and policies on real robots, demonstrating continuous control for T-shirt grasping.
Visual Language Maps for Robot Navigation
Chen Huang, Wolfram Burgard
Robotic IntelligenceLarge Language ModelVision Language ModelSimultaneous Localization and MappingImageTextMultimodality
🎯 What it does: Propose a spatial map representation method called VLMaps, which combines pre-trained vision-language features with robot 3D reconstruction to achieve automated map construction and support natural language indexing.
Visual Pitch and Roll Estimation For Inland Water Vessels
Dennis Grießer, M. Franz
SegmentationPose EstimationConvolutional Neural NetworkImage
🎯 What it does: A visual pitch and roll angle estimation method for inland waterway vessels was developed, utilizing CNN for water surface segmentation, stereo vision reconstruction, and geometric calculations to estimate pitch and roll.
Visual Tracking of Needle Tip in 2D Ultrasound based on Global Features in a Siamese Architecture
Wanquan Yan, S. Cheng
Object TrackingBiomedical DataUltrasound
🎯 What it does: Proposed a learning-based visual tracking network based on the Siamese structure for tracking the tip position in ultrasound images.
Visual-Inertial and Leg Odometry Fusion for Dynamic Locomotion
Victor Dhédin, Joerg Stueckler
Robotic IntelligenceSimultaneous Localization and MappingMultimodality
🎯 What it does: Propose an EKF state estimator that integrates Visual-Inertial Odometry (VIO) with leg odometry, and embeds it into a nonlinear model predictive controller (NMPC) to achieve accurate pose estimation and control for quadruped robots during dynamic motions such as jumping and lateral running.
Visuomotor Control in Multi-Object Scenes Using Object-Aware Representations
Negin Heravi, Debidatta Dwibedi
Robotic Intelligence
🎯 What it does: Using self-supervised, object-aware representation learning methods to learn robot control policies and object localization in multi-object scenes
Vitreoretinal Surgical Robotic System with Autonomous Orbital Manipulation using Vector-Field Inequalities
Yuki Koyama, K. Harada
Robotic Intelligence
🎯 What it does: Proposes a method for autonomous ocular orbit manipulation in robot-assisted vitreoretinal surgery, utilizing a remote control system to achieve eye rotation for expanding the retinal field of view
VP-STO: Via-point-based Stochastic Trajectory Optimization for Reactive Robot Behavior
Julius Jankowski, S. Calinon
OptimizationRobotic Intelligence
🎯 What it does: Proposed a VP-STO framework based on virtual points for jointly optimizing smooth and time-optimal robot trajectories in both space and time, achieving real-time control in simulation and real-robot closed-loop operations.
VQA-based Robotic State Recognition Optimized with Genetic Algorithm
Kento Kawaharazuka, M. Inaba
RecognitionOptimizationRobotic IntelligenceTransformerVision Language ModelMultimodality
🎯 What it does: Using pre-trained vision-language models (PTVLM) combined with visual question answering (VQA) methods in robot state recognition, and optimizing the question combinations through genetic algorithms to enhance recognition performance.
WAVN: Wide Area Visual Navigation for Large-scale, GPS-denied Environments
D. Lyons, Mohamed Rahouti
Robotic IntelligenceSimultaneous Localization and MappingImage
🎯 What it does: Proposes a multi-robot visual navigation method called WAVN for large-scale, GPS-denied environments, utilizing the visual homing paradigm and team-wide visual information to navigate to distant targets.
Wayformer: Motion Forecasting via Simple & Efficient Attention Networks
Nigamaa Nayakanti, Benjamin Sapp
Autonomous DrivingComputational EfficiencyTransformer
🎯 What it does: Propose a concise and unified architecture called Wayformer based on attention mechanisms for motion prediction, and investigate different modality fusion strategies
WE-Filter: Adaptive Acceptance Criteria for Filter-based Shared Autonomy
Michael Bowman, Xiaoli Zhang
Robotic Intelligence
🎯 What it does: Proposed an adaptive acceptance criterion named WE-Filter for filtered shared autonomous control.
Weakly Supervised Referring Expression Grounding via Target-Guided Knowledge Distillation
Jinpeng Mi, Jianwei Zhang
RecognitionKnowledge DistillationImageTextMultimodalityBenchmark
🎯 What it does: Proposes a target-guided knowledge distillation framework for weakly supervised referring expression localization, leveraging target-related prediction information from a pre-trained teacher model to guide student model training, thereby enhancing weakly supervised localization performance.
WEDGE: Web-Image Assisted Domain Generalization for Semantic Segmentation
Nam-Won Kim, Suha Kwak
SegmentationDomain AdaptationImage
🎯 What it does: Proposed the WEDGE scheme, leveraging diverse images crawled from the web to enhance the domain generalization capability of semantic segmentation.
Weighted Maximum Likelihood for Controller Tuning
Angel Romero, D. Scaramuzza
Autonomous DrivingOptimization
🎯 What it does: Automatically learn the optimal objective function for MPCC using a weighted maximum likelihood (WML) method, enhancing control performance in drone path tracking and progress balancing.
Wi-Closure: Reliable and Efficient Search of Inter-robot Loop Closures Using Wireless Sensing
Weiying Wang, Stephanie Gil
OptimizationComputational EfficiencyRobotic IntelligenceSimultaneous Localization and Mapping
🎯 What it does: Proposed the Wi-Closure algorithm to enhance computational efficiency and robustness in loop closure detection for multi-robot SLAM.
Wide-Area Geolocalization with a Limited Field of View Camera
Lena M. Downes, J. How
Convolutional Neural NetworkSimultaneous Localization and MappingImage
🎯 What it does: Proposes a cross-view geolocation method called ReWAG for standard non-panoramic ground cameras, leveraging a Siamese network combining pose-aware embeddings and particle filters to achieve global localization in GPS-denied environments using only odometry and 90° field-of-view camera data.
Wild-Places: A Large-Scale Dataset for Lidar Place Recognition in Unstructured Natural Environments
Joshua Knights, P. Moghadam
RecognitionPoint Cloud
🎯 What it does: Proposed and released a large-scale LiDAR place recognition dataset called Wild-Places, focusing on unstructured natural environments, and providing multiple revisits with 6DoF ground truth;
Wirelessly-Controlled Untethered Piezoelectric Planar Soft Robot Capable of Bidirectional Crawling and Rotation
Zhiwu Zheng, J. Sturm
Robotic Intelligence
🎯 What it does: Designed and demonstrated a wireless, cordless planar soft robot with five piezoelectric actuators capable of bidirectional crawling, turning, and in-place rotation;
WorldGen: A Large Scale Generative Simulator
Chahat Deep Singh, Y. Aloimonos
GenerationData SynthesisGenerative Adversarial NetworkPoint CloudMesh
🎯 What it does: Proposed WorldGen, an open-source framework capable of automatically generating a large number of structured and unstructured 3D realistic scenes along with their complete annotations.
WS-3D-Lane: Weakly Supervised 3D Lane Detection With 2D Lane Labels
Jianyong Ai, Jiachen Zhong
Object DetectionAutonomous DrivingImage
🎯 What it does: Propose a weakly supervised 3D lane detection method called WS-3D-Lane, which trains 3D lanes using only 2D lane labels.
Zero-shot Active Visual Search (ZAVIS): Intelligent Object Search for Robotic Assistants
Jeongeun Park, Sungjoon Choi
Robotic IntelligenceVision Language ModelVision-Language-Action ModelSimultaneous Localization and MappingImageTextMultimodality
🎯 What it does: Proposed a zero-shot active visual search system called ZAVIS, which uses a mobile robot and visual sensors to locate target objects through free-text commands;
Zero-shot Object Detection Based on Dynamic Semantic Vectors
Haoyu Li, Yu Hu
Object DetectionAutonomous DrivingRepresentation LearningContrastive LearningImage
🎯 What it does: Proposes a zero-shot object detection method based on dynamic semantic vectors, and designs a bidirectional classification branch network with an optimization process incorporating the N-pair loss
Zero-Shot Object Goal Visual Navigation
Qianfan Zhao, Zhi-yong Liu
Object DetectionAutonomous DrivingRepresentation LearningVision-Language-Action ModelImageText
🎯 What it does: Studied the zero-shot goal-oriented visual navigation task and proposed the Semantic Similarity Network (SSNet) framework;
Zero-Shot Policy Transfer with Disentangled Task Representation of Meta-Reinforcement Learning
Zheng Wu, M. Tomizuka
Representation LearningMeta LearningReinforcement Learning
🎯 What it does: Proposed a meta-reinforcement learning algorithm with decoupled task representations to achieve zero-shot policy generalization for RL agents on unseen composite tasks.
Zero-Shot Transfer of Haptics-Based Object Insertion Policies
Samarth Brahmbhatt, M. Muller
Domain AdaptationRobotic Intelligence
🎯 What it does: Trained a tactile-driven object insertion strategy learned in simulation for placing plates into a grooved tray, achieving zero-shot transfer to a real robot without fine-tuning.