arXivSub Start free trial

ICRA 2023 Papers — Page 10

IEEE International Conference on Robotics and Automation · 1341 papers

Online Visual SLAM Adaptation against Catastrophic Forgetting with Cycle-Consistent Contrastive Learning

Sangni Xu, Zhiyong Wang

Image TranslationDomain AdaptationContrastive LearningSimultaneous Localization and MappingImage

🎯 What it does: Propose an online self-supervised adaptation method that continuously updates the pre-trained visual SLAM model using a novel adapter to cope with environmental changes.

Online Whole-Body Motion Planning for Quadrotor using Multi-Resolution Search

Yunfan Ren, Fu Zhang

Robotic IntelligencePoint Cloud

🎯 What it does: Proposes a multi-resolution search method for UAV whole-body motion planning in unknown unstructured environments, automatically decomposing the problem into SE(3) and R3 sub-problems, and designing a passage generation strategy for narrow areas to improve success rates.

Open-vocabulary Queryable Scene Representations for Real World Planning

Boyuan Chen, Daniel Kappler

Representation LearningRobotic IntelligenceTransformerLarge Language ModelVision Language ModelMultimodality

🎯 What it does: Propose NLMap, an open-vocabulary and queryable scene representation framework, enabling the LLM planner to query the availability and location of objects in the scene before generating a plan, thus achieving context-aware robot planning.

Operating with Inaccurate Models by Integrating Control-Level Discrepancy Information into Planning

Ellis Ratner, M. Likhachev

OptimizationRobotic IntelligenceWorld Model

🎯 What it does: By introducing statistical analysis of control layer model errors during the planning phase, the error distribution is inferred and converted into a cost bias in the planning layer, enabling the robot to complete tasks more efficiently even under model inaccuracies.

Operative Action Captioning for Estimating System Actions

Taiki Nakamura, Koichiro Yoshino

GenerationVision-Language-Action ModelImageText

🎯 What it does: Proposed an operative action captioning task and constructed a system that generates textual descriptions of required operations based on two images of the current and target states;

OPT-Mimic: Imitation of Optimized Trajectories for Dynamic Quadruped Behaviors

Yuni Fuchioka, M. V. D. Panne

Robotic IntelligenceReinforcement LearningWorld Model

🎯 What it does: Imitate reference trajectories generated by model-based optimization using reinforcement learning (RL) to train four dynamic quadruped robot behaviors (walking, forward jump, 180-degree backward roll, bipedal walking), and directly transfer the trained policies from simulation to the real Solo 8 robot;

OpTaS: An Optimization-based Task Specification Library for Trajectory Optimization and Model Predictive Control

Christopher E. Mower, Christos Bergeles

OptimizationRobotic Intelligence

🎯 What it does: Presents OpTaS, a Python task specification library for robot trajectory optimization and model predictive control.

OptiGap: A Modular Optical Sensor System for Bend Localization

Paul Bupe, C. Harnett

ClassificationRecognition

🎯 What it does: Studied an optical sensor that utilizes air gaps in flexible light guides to form encoding patterns for bending positioning.

Optimal Allocation of Many Robot Guards for Sweep-Line Coverage

S. Feng, Jingjin Yu

OptimizationRobotic Intelligence

🎯 What it does: A study and proposal of an optimal algorithm for the multi-robot allocation problem in known 2D environments with predefined sweeping plans, aiming to achieve desired coverage quality through time-parameterized robot distribution or maximize probability guarantees given a fixed number of robots.

Optimal Grasps and Placements for Task and Motion Planning in Clutter

Carlos Quintero-Peña, L. Kavraki

OptimizationRobotic Intelligence

🎯 What it does: Proposed an optimization-based grounding method to address grasping and placement problems in task and motion planning under cluttered environments.

Optimal Multi-Robot Coverage Path Planning for Agricultural Fields using Motion Dynamics

J. Choton, P. Prabhakar

OptimizationRobotic IntelligenceAgriculture Related

🎯 What it does: Proposes a multi-robot agricultural field coverage path planning method that utilizes motion dynamics and minimizes task time

Optimal Parameterized Joints Selection to Improve Motion Planning Performance of Redundant Manipulators

Bin Xie, Di Wu

OptimizationRobotic Intelligence

🎯 What it does: Proposed and verified a method for selecting optimal parameterized joints to enhance the motion planning performance of redundant robotic arms, with experimental validation conducted on an 8-degree-of-freedom (DOF) robotic arm.

Optimal Scheduling of Models and Horizons for Model Hierarchy Predictive Control

Charles Khazoom, Sangbae Kim

Optimization

🎯 What it does: Propose a framework that systematically optimizes model scheduling in model hierarchical predictive control (MHPC) by approximating the closed-loop cost through trajectory optimization, reducing the number of decision variables while maintaining performance.

Optimal Workpiece Placement Based on Robot Reach, Manipulability and Joint Torques

Baris Balci, Peter Corke

OptimizationRobotic Intelligence

🎯 What it does: Proposes a nonlinear optimization-based algorithm to determine the optimal workpiece pose, enabling the robot to achieve minimal joint torque and maximal manipulability during surface polishing.

Optimized Design and Analysis of Active Propeller-driven Capsule Endoscopic Robot for Gastric Examination

Yi Zhang, Chengzhi Hu

OptimizationRobotic IntelligenceBiomedical Data

🎯 What it does: Proposed an integrated propulsion capsule endoscope robot (PCER) and verified its mobility in a liquid environment through finite element analysis, CFD simulation, and experiments.

Optimizing Bipedal Locomotion for The 100m Dash With Comparison to Human Running

Devin Crowley, Alan Fern

OptimizationRobotic Intelligence

🎯 What it does: Explore the gait space of the Cassie robot at different speeds, propose methods to optimize gait efficiency, and integrate them into a complete controller to complete a 100-meter race

Option-Aware Adversarial Inverse Reinforcement Learning for Robotic Control

Jiayu Chen, V. Aggarwal

Robotic IntelligenceReinforcement LearningAuto EncoderGenerative Adversarial Network

🎯 What it does: Proposes a hierarchical imitation learning algorithm based on adversarial inverse reinforcement learning, which directly recovers hierarchical policies from unannotated demonstrations using the EM algorithm, introduces directional information terms to enhance causal relationships, and employs variational autoencoders (VAEs) to achieve end-to-end learning.

Orbeez-SLAM: A Real-time Monocular Visual SLAM with ORB Features and NeRF-realized Mapping

Chi-Ming Chung, Winston H. Hsu

Neural Radiance FieldSimultaneous Localization and MappingImage

🎯 What it does: Developed a real-time monocular visual SLAM called Orbeez-SLAM, which combines ORB features and NeRF to achieve dense map mapping, enabling rapid adaptation to new scenes without pre-training and generating real-time dense maps.

Origami Folding Enhances Modularity and Mechanical Efficiency of Soft Actuators

Zheng Wang, Hongying Zhang

Robotic Intelligence

🎯 What it does: Designed, prototyped, and tested three types of modular origami soft actuators capable of generating translational, bending, and twisting movements

ORORA: Outlier-Robust Radar Odometry

Hyungtae Lim, H. Myung

Pose EstimationAutonomous DrivingOptimizationPoint Cloud

🎯 What it does: Proposed a robust outlier detection and estimation method called ORORA for radar odometry.

Output Mode Switching for Parallel Five-bar Manipulators Using a Graph-based Path Planner

Parker B. Edwards, J. Hauenstein

OptimizationRobotic IntelligenceGraph

🎯 What it does: Model the configuration space of a parallel five-bar mechanism using a radius graph, assign weights to each edge via a homotopy continuation optimization method, and then use a graph path planner to find approximate geodesics that avoid low transmission quality regions, automatically generating motion paths capable of switching between non-adjacent output modes.

OysterNet: Enhanced Oyster Detection Using Simulation

Xiao-sheng Lin, Y. Aloimonos

Object DetectionData SynthesisImage

🎯 What it does: Propose a method that mathematically models sea snails and renders images in simulations, using synthetic data to enhance the detection performance of the OysterNet network.

Parallel Inversion of Neural Radiance Fields for Robust Pose Estimation

Yunzhi Lin, Stan Birchfield

Pose EstimationOptimizationNeural Radiance FieldImage

🎯 What it does: Propose a parallelized optimization method based on fast NeRF, which estimates the camera's 6-DoF pose by minimizing the residual between rendered pixels and observed pixels using a single RGB image.

Parallel Reinforcement Learning Simulation for Visual Quadrotor Navigation

Jack D. Saunders, Wenbin Li

Computational EfficiencyRobotic IntelligenceReinforcement Learning

🎯 What it does: Proposed a parallel reinforcement learning simulation framework based on AirSim, and modified Ape-X to achieve multi-machine parallel training;

Parameter Optimization for Manipulator Motion Planning using a Novel Benchmark Set

Carl Gäbert, Ulrike Thomas

OptimizationHyperparameter SearchRobotic IntelligenceBenchmark

🎯 What it does: Analyzed parameter settings of optimization planners for a large number of diverse planning problems, explored the relationship between problem characteristics and optimal parameters, and provided recommended parameter lists for different application scenarios.

Parameter-Conditioned Reachable Sets for Updating Safety Assurances Online

Javier Borquez, Somil Bansal

Safty and Privacy

🎯 What it does: Propose a parameterized conditional reachable set method to achieve online safety guarantee updates

PARSEC: An Aerial Platform for Autonomous Deployment of Self-Anchoring Payloads on Natural Vertical Surfaces

Patrick Spieler, J. Burdick

Robotic Intelligence

🎯 What it does: Developed an aerial operation platform called PARSEC capable of autonomously deploying self-anchoring payloads, utilizing a six-rotor helicopter and a two-degree-of-freedom mass-balanced robotic arm to achieve self-anchoring of payloads on vertical rock surfaces using micro claws, while wirelessly transmitting sensor data back to the host computer.

Passive robotic gripper using a contact-based locking mechanism

Issei Nate, S. Hirai

Robotic Intelligence

🎯 What it does: A passive mechanical gripper without actuators was proposed and fabricated, employing a contact-based locking mechanism. Grasping shape modeling and experimental validation were conducted, and the gripper's ability to grasp objects of different sizes, weights, and various food items was tested.

Passivity-based Decentralized Control for Collaborative Grasping of Under-Actuated Aerial Manipulators

Jinyeong Jeong, Min Jun Kim

Robotic Intelligence

🎯 What it does: Proposed a decentralized control scheme based on passive impedance control for cooperative grasping of underactuated aerial manipulators.

Path Planning Under Uncertainty to Localize mmWave Sources

K. Pfeiffer, L. Righetti

OptimizationRobotic Intelligence

🎯 What it does: Developed an algorithm for signal estimation and path planning, helping mobile robots locate millimeter-wave wireless signals in crowded indoor environments by leveraging the directional characteristics of signals combined with Extended Kalman Filter (EKF) and belief space dynamics for path planning.

PCGen: Point Cloud Generator for LiDAR Simulation

Chenqi Li, Bingbing Liu

Data SynthesisAutonomous DrivingPoint Cloud

🎯 What it does: Proposed PCGen, a point cloud generator for LiDAR simulation, addressing the issues of traditional methods producing overly complete and less noisy point clouds, as well as insufficient simulation for non-rigid objects.

Pedestrian Crossing Action Recognition and Trajectory Prediction with 3D Human Keypoints

Jiachen Li, Congcong Li

RecognitionPose EstimationContrastive Learning

🎯 What it does: This paper extracts 3D human keypoints from raw sensor data and constructs a multi-task learning framework to achieve pedestrian crossing behavior recognition and trajectory prediction;

PedFormer: Pedestrian Behavior Prediction via Cross-Modal Attention Modulation and Gated Multitask Learning

Amir Rasouli, Iuliia Kotseruba

Autonomous DrivingTransformerVideo

🎯 What it does: Propose a pedestrian behavior prediction framework based on a multi-modal cross-attention transformer.

Perceiving Unseen 3D Objects by Poking the Objects

Linghao Chen, Xiaowei Zhou

Robotic IntelligenceSupervised Fine-Tuning

🎯 What it does: Utilizing gesture-based touching for interactive perception, enabling robots to automatically discover and reconstruct unseen 3D objects

Performance Evaluation of 3D Keypoint Detectors and Descriptors on Coloured Point Clouds in Subsea Environments

Kyung-Tae Jung, J. Forbes

Pose EstimationMultimodalityPoint Cloud

🎯 What it does: Studied 3D keypoint detectors and feature descriptors using colored point clouds in underwater environments, and evaluated their performance using a dataset of field-collected underwater laser scans.

Perturbation-Based Best Arm Identification for Efficient Task Planning with Monte-Carlo Tree Search

Daejong Jin, Kyungjae Lee

OptimizationReinforcement Learning

🎯 What it does: Propose a tree search method based on perturbation-driven optimal arm identification (PBAI) to improve Monte Carlo Tree Search (MCTS) in task and motion planning, achieving better balance between exploration and exploitation while accelerating discovery of globally optimal plans.

Pick2Place: Task-aware 6DoF Grasp Estimation via Object-Centric Perspective Affordance

Zhanpeng He, Volkan Isler

Pose EstimationOptimizationRobotic Intelligence

🎯 What it does: Propose a task-aware 6DoF grasp estimation method that integrates grasp and placement coordination, generating placement affinity maps using an object-centric action space.

Picking by Tilting: In-Hand Manipulation for Object Picking using Effector with Curved Form

Yanshu Song, Yun-hui Liu

Robotic Intelligence

🎯 What it does: Proposed a robotic hand operation technique using a curved passive end-effector and two plane support surfaces for tilted grasping to handle objects that are too large for conventional grasping.

PIEKF-VIWO: Visual-Inertial-Wheel Odometry using Partial Invariant Extended Kalman Filter

Tong Hua, L. Pei

Autonomous DrivingOptimizationSimultaneous Localization and MappingMultimodality

🎯 What it does: Proposes a Partial Invariant Extended Kalman Filter (PIEKF) that incorporates only the rotation-speed state into the Lie group structure, and applies it to Visual-Inertial-Odometry with Wheel Speed (VIWO) to improve localization accuracy and consistency.

Place Recognition under Occlusion and Changing Appearance via Disentangled Representations

Yue Chen, Xingyu Chen

RecognitionImage

🎯 What it does: Propose an unsupervised method called PROCA that decomposes image representations into scene codes, appearance codes, and occlusion codes for scene recognition under occlusions and appearance variations.

Planning Assembly Sequence with Graph Transformer

Lin Ma, Guyue Zhou

OptimizationGraph Neural NetworkTransformerGraphBenchmark

🎯 What it does: This paper proposes a framework for assembly sequence planning based on graph Transformer, which is trained and validated on a heterogeneous graph dataset composed of self-collected LEGO models;

Planning for Complex Non-prehensile Manipulation Among Movable Objects by Interleaving Multi-Agent Pathfinding and Physics-Based Simulation

D. Saxena, M. Likhachev

Robotic Intelligence

🎯 What it does: A planning method for non-grasping rearrangement among movable objects in cluttered environments was developed, which evaluates complex robot-object and object-object interactions through physics simulation.

Planning for Multi-Object Manipulation with Graph Neural Network Relational Classifiers

Yixuan Huang, Tucker Hermans

Robotic IntelligenceGraph Neural NetworkPoint Cloud

🎯 What it does: Developed a new framework based on graph neural networks for multi-object manipulation, capable of predicting changes in object relationships caused by robot actions and achieving multi-step planning to achieve target relationships; the framework operates on partial viewpoint point clouds and can realize object reorganization in real environments.

Planning with Occluded Traffic Agents using Bi-Level Variational Occlusion Models

Filippos Christianos, M. Pavone

GenerationAutonomous DrivingPoint Cloud

🎯 What it does: Proposed and implemented a two-step generative model called BiVO, which first predicts the position of occluded traffic participants and then generates their potential trajectories, using the trajectory distribution for downstream planning; meanwhile, conducted closed-loop replanning simulation evaluation on the nuScenes dataset.

Planning with SiMBA: Motion Planning under Uncertainty for Temporal Goals using Simplified Belief Guides

Qi Heng Ho, Morteza Lahijanian

Autonomous Driving

🎯 What it does: Designed a multi-layer algorithm that employs a trajectory-guided sampling search tree using simplified models in task and belief spaces to achieve motion planning under motion and perception uncertainties for linear temporal logic (LTL) specifications.

PointCloudLab: An Environment for 3D Point Cloud Annotation with Adapted Visual Aids and Levels of Immersion

Achref Doula, Alejandro Sánchez Guinea

Point Cloud

🎯 What it does: Proposes PointCloudLab, a 3D point cloud annotation environment that supports multiple levels of immersion and visualization aids, and validates its effectiveness through controlled experiments.

Policy-Guided Lazy Search with Feedback for Task and Motion Planning

M. Khodeir, Florian Shkurti

Computational EfficiencyRobotic Intelligence

🎯 What it does: Proposed a PDDLStream solver named LAZY that combines action skeleton planning with lazy geometric sampling in a single integrated search, and adaptively guides task planning using learned goal-directed policies and current motion sampling data.

Portable Multi-Hypothesis Monte Carlo Localization for Mobile Robots

Alberto García, Vicente Matellán Olivera

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Propose a multi-hypothesis Monte Carlo localization algorithm that maintains multiple particle clusters and selects the optimal one as the output.

Pose Quality Prediction for Vision Guided Robotic Shoulder Arthroplasty

Morgan Windsor, Michael Milford

Robotic IntelligenceImageBiomedical Data

🎯 What it does: Propose a lightweight method that utilizes internal pipeline products to predict the performance of visual localization estimation, and demonstrate that online performance prediction can drive robot navigation, significantly reducing localization error.

Pose Relation Transformer Refine Occlusions for Human Pose Estimation

Hyung-gun Chi, K. Ramani

Pose EstimationTransformer

🎯 What it does: Proposed a module named POse Relation Transformer (PORT) for correcting missing keypoints caused by occlusion in human pose estimation.

Pose-graph SLAM Using Multi-order Ultrasonic Echoes and Beamforming for Long-range Inspection Robots

Othmane-Latif Ouabi, Cédric Pradalier

OptimizationRobotic IntelligenceSimultaneous Localization and MappingUltrasound

🎯 What it does: Proposed a graph SLAM method based on ultrasonic guided wave multi-stage echoes and beamforming for long-range metal plate structure inspection tasks.

Practical Visual Deep Imitation Learning via Task-Level Domain Consistency

Mohi Khansari, Eric Jang

Domain AdaptationRobotic IntelligenceGenerative Adversarial NetworkImage

🎯 What it does: This paper proposes a visual depth imitation learning method based on task-level domain consistency, utilizing RGB and depth images to control a 9-degree-of-freedom mobile robot to complete the task of opening a latched door.

Predicting Motion Plans for Articulating Everyday Objects

Arjun Gupta, Saurabh Gupta

Data SynthesisRobotic Intelligence

🎯 What it does: Developed a learning-based framework for directly predicting joint motion plans in new environments for mobility manipulation tasks (e.g., opening doors, pulling out drawers, lifting toilet seats), generating simulated data for articulated objects in real-world scenarios through the ArtObjSim simulator; adopted SeqIK+θ0 as a fast and flexible motion planning representation; trained the model to rapidly generate motion plans for novel objects.

Predictive Runtime Verification of Skill-based Robotic Systems using Petri Nets

Baptiste Pelletier, M. Rognant

Robotic Intelligence

🎯 What it does: Using Petri nets for online supervision of skill-based robotic systems composed of multiple complex components.

PredRecon: A Prediction-boosted Planning Framework for Fast and High-quality Autonomous Aerial Reconstruction

Chen Feng, S. Shen

OptimizationRobotic IntelligenceImagePoint Cloud

🎯 What it does: Propose the PredRecon framework, which utilizes autonomous path generation by drones to achieve fast and high-quality 3D reconstruction.

Preliminary Evaluation of a Wearable Thruster for Arresting Backwards Falls

M. Henry, M. Goldfarb

Safty and PrivacyRobotic Intelligence

🎯 What it does: Integrate a nitrogen-based cold gas thruster into a backpack prototype device and conduct experiments on three healthy subjects to evaluate its effect on backward falls

Prioritized Robotic Exploration with Deadlines: A Comparison of Greedy, Orienteering, and Profitable Tour Approaches

S. Datta, Srinivas Akella

OptimizationRobotic IntelligenceGraph

🎯 What it does: Studied the problem of priority exploration by mobile robots in unknown indoor environments with a known deadline, aiming to quickly generate a geometric layout and return to the starting point within the deadline; modeled priority exploration as the Orienteering Problem (OP) and Profitable Tour Problem (PTP), and compared a priority-based greedy algorithm with OP/PTP algorithms; conducted experiments in various graph-based and Gazebo simulation environments.

PriorLane: A Prior Knowledge Enhanced Lane Detection Approach Based on Transformer

Qibo Qiu, Xiaofei He

SegmentationAutonomous DrivingTransformerImage

🎯 What it does: Propose the PriorLane framework, which utilizes an encoder-only Transformer to integrate features from pre-trained segmentation models with low-cost local prior knowledge embeddings, thereby enhancing lane detection segmentation performance.

Privacy-Preserving Video Conferencing via Thermal-Generative Images

Sheng-Yang Chiu, Yu-Chin Nieh

Image TranslationGenerationSafty and PrivacyConvolutional Neural NetworkGenerative Adversarial NetworkImageMultimodality

🎯 What it does: Using low-resolution thermal imaging maps as conditions to guide RGB image synthesis, thus achieving privacy-preserving video conferencing.

Probabilistic Contact State Estimation for Legged Robots using Inertial Information

Michael Maravgakis, P. Trahanias

Robotic IntelligenceTime Series

🎯 What it does: Estimate the contact state between the robot's foot and the ground using only proprioceptive data from the robot's end-effector IMU through probabilistic methods.

Probabilistic Plane Extraction and Modeling for Active Visual-Inertial Mapping

Mitchell Usayiwevu, Teresa Vidal-Calleja

Convolutional Neural NetworkSimultaneous Localization and MappingImage

🎯 What it does: Proposes an active visual-inertial mapping framework that combines point features with probabilistic plane extraction and modeling, and achieves delayed processing through IEKF and measurement extrapolation; simultaneously employs information path planning to realize active map construction.

Probabilistic Planning with Partially Ordered Preferences over Temporal Goals

Hazhar Rahmani, Jie Fu

OptimizationReinforcement Learning

🎯 What it does: Proposes a method for temporal goal planning in Markov Decision Processes (MDP) under partial order preferences (non-total order), and provides the corresponding planning algorithm.

Probabilistic Rare-Event Verification for Temporal Logic Robot Tasks

Guy Scher, H. Kress-Gazit

Robotic IntelligenceFlow-based Model

🎯 What it does: Proposes a method for calculating the probability of a robot successfully completing tasks described by Signal Temporal Logic (STL), particularly focusing on rare events with extremely low failure probabilities.

Probabilistic Risk Assessment for Chance-Constrained Collision Avoidance in Uncertain Dynamic Environments

Khaled A. Mustafa, Javier Alonso-Mora

Autonomous DrivingOptimization

🎯 What it does: Propose a real-time feasible method to quantify the risk of planning trajectories from various probabilistic planners (with different risk upper limits), and select the least conservative trajectory with risk below the threshold to achieve a balance between safety and efficiency;

Probabilistic Uncertainty Quantification of Prediction Models with Application to Visual Localization

Junan Chen, Mark E. Campbell

Autonomous DrivingSimultaneous Localization and MappingVideo

🎯 What it does: Propose a method that utilizes a sensor error model to map the internal output of a predictive model to probabilistic uncertainty in visual localization, and incorporates this uncertainty into a Kalman filter for localization;

Proficiency Self-Assessment without Breaking the Robot: Anomaly Detection using Assumption-Alignment Tracking from Safe Experiments

Xuan Cao, M. Goodrich

Anomaly DetectionRobotic Intelligence

🎯 What it does: Proposes an anomaly detection method based on a one-class classifier, utilizing feature vectors generated by Assumption Consistency Tracking (AAT). It identifies clusters under different operating conditions through two metrics, difference and separability, enabling anomaly detection using only normal data. The method's effectiveness is validated in simulation navigation robot experiments and Sawyer robot grasping tasks.

ProgPrompt: Generating Situated Robot Task Plans using Large Language Models

Ishika Singh, Animesh Garg

Robotic IntelligenceLarge Language ModelPrompt EngineeringText

🎯 What it does: Proposed a programmatic LLM prompt structure for generating applicable robot operation plans across various environments, robot capabilities, and tasks.

Proprioceptive Sensor-Based Simultaneous Multi-Contact Point Localization and Force Identification for Robotic Arms

Seohee Han, Min Jun Kim

Computational EfficiencyRobotic IntelligenceMesh

🎯 What it does: Propose a self-perception algorithm for a robotic arm using joint torque sensors and base force/torque sensors, capable of simultaneously estimating contact point location and contact force

Puppeteer and Marionette: Learning Anticipatory Quadrupedal Locomotion Based on Interactions of a Central Pattern Generator and Supraspinal Drive

M. Shafiee, A. Ijspeert

Robotic IntelligenceReinforcement Learning

🎯 What it does: Studied the interaction between supraspinal drive forces and the spinal central pattern generator (CPG) during anticipatory quadrupedal gap crossing, and utilized deep reinforcement learning to train neural network strategies that simulate supraspinal drive behavior. Quadruped robots can be controlled by adjusting CPG dynamics or directly modifying execution signals to cross gaps.

Pyramid Learnable Tokens for 3D LiDAR Place Recognition

Congcong Wen, Yi Fang

RecognitionAutonomous DrivingTransformerPoint Cloud

🎯 What it does: Propose a point transformation network called PTNet-PLT, using pyramid learnable tokens to learn global descriptors for 3D LiDAR real-world scans, for place recognition.

QuadMag: A Mobile-Coil System With Enhanced Magnetic Actuation Efficiency and Dexterity

Lidong Yang, Li Zhang

Robotic IntelligencePhysics Related

🎯 What it does: Propose the QuadMag four-coil moving coil system and design its motion control based on a parallel mechanism.

Quadruped Guidance Robot for the Visually Impaired: A Comfort-Based Approach

Yanbo Chen, Bin Liang

Robotic Intelligence

🎯 What it does: Proposed a vision-based obstacle assistance navigation system for a quadruped robot (Unitree Laikago), comprising an adjustable-length and force traction device, motion planning driven by a human-robot mechanics model, and corresponding robot motion planning and force control mechanisms, which have been validated in real environments.

Question Generation for Uncertainty Elimination in Referring Expressions in 3D Environments

F. Matsuzawa, Y. Satoh

RecognitionGenerationMultimodality

🎯 What it does: Propose a method to eliminate uncertainty in referential expressions in 3D indoor environments through questioning, and construct a corresponding dataset.

Radar Velocity Transformer: Single-scan Moving Object Segmentation in Noisy Radar Point Clouds

M. Zeller, C. Stachniss

SegmentationTransformerPoint Cloud

🎯 What it does: Achieving moving object segmentation in noisy radar point clouds using a single scan

RAMP-Net: A Robust Adaptive MPC for Quadrotors via Physics-informed Neural Network

Sourav Sanyal, Kaushik Roy

OptimizationRobotic IntelligencePhysics RelatedOrdinary Differential Equation

🎯 What it does: Proposed a robust adaptive MPC framework called RAMP-Net based on physics-informed neural networks for quadrotor trajectory tracking.

RAMP: A Risk-Aware Mapping and Planning Pipeline for Fast Off-Road Ground Robot Navigation

Lakshay Sharma, J. How

Robotic IntelligenceSimultaneous Localization and Mapping

🎯 What it does: Proposed the RAMP pipeline, improving mapping and planning to achieve fast and safe ground robot navigation

RAMP: Reaction-Aware Motion Planning of Multi-Legged Robots for Locomotion in Microgravity

Warley F. R. Ribeiro, Kazuya Yoshida

Robotic Intelligence

🎯 What it does: Proposed a reaction-aware motion planning (RAMP) for multi-legged robot locomotion in microgravity environments, minimizing swing momentum through low-reactive swing trajectory (LRST) and distributing momentum across the entire body to ensure zero velocity at supporting grippers, thereby reducing the risk of grippers detaching from the ground.

RangedIK: An Optimization-based Robot Motion Generation Method for Ranged-Goal Tasks

Yeping Wang, Michael Gleicher

OptimizationRobotic Intelligence

🎯 What it does: Propose a real-time motion generation method that unifies the handling of multi-task scenarios with specific goals, equivalent goal ranges, and priority goal ranges, leveraging task flexibility within goal ranges to satisfy other tasks.

ReachLipBnB: A branch-and-bound method for reachability analysis of neural autonomous systems using Lipschitz bounds

Taha Entesari, Mahyar Fazlyab

Optimization

🎯 What it does: Proposes a neural network reachability analysis method based on branch-and-bound, applicable to both open-loop and closed-loop control systems.

Real World Offline Reinforcement Learning with Realistic Data Source

G. Zhou, Vikash Kumar

Robotic IntelligenceReinforcement LearningTime Series

🎯 What it does: Evaluate the generalization and transferability of offline reinforcement learning methods on four real-world desktop manipulation tasks, collecting 6,500+ trajectories, 800+ robot hours, and 270+ human hours of data.

Real-time Acoustic Holography with Iterative Unsupervised Learning for Acoustic Robotic Manipulation

Chengxi Zhong, Song Liu

Robotic IntelligenceAudio

🎯 What it does: Propose a real-time phase-only acoustic holography algorithm that utilizes iterative unsupervised learning combined with an experience pool constructed using a physical model to achieve mapping between the target acoustic amplitude hologram and POH.

Real-time Background Subtraction under Varying Lighting Conditions

Sisi Liang, D. Baker

SegmentationVideoBenchmark

🎯 What it does: Designed and implemented the EGMM method for real-time background subtraction, effectively handling illumination change issues.

Real-Time Constrained 6D Object-Pose Tracking of An In-Hand Suture Needle for Minimally Invasive Robotic Surgery

Zih-Yun Chiu, Michael C. Yip

Pose EstimationRobotic Intelligence

🎯 What it does: Real-time 6D pose tracking of a needle inside the hand was achieved using Bayesian filtering with feasible grasp constraints.

Real-Time Decentralized Navigation of Nonholonomic Agents Using Shifted Yielding Areas

Liang He, Dinesh Manocha

OptimizationRobotic Intelligence

🎯 What it does: Proposed a lightweight decentralized algorithm for navigating multiple heterogeneous robots in challenging environments with narrow passages.

Real-Time Dense 3D Mapping of Underwater Environments

Weihan Wang, Ioannis M. Rekleitis

Robotic IntelligenceSimultaneous Localization and MappingImage

🎯 What it does: Proposed a real-time dense 3D reconstruction pipeline suitable for resource-constrained underwater unmanned vehicles, integrating SVIn2 visual-inertial odometry to achieve online reconstruction.

Real-Time Estimation of Walking Speed and Stride Length Using an IMU Embedded in a Robotic Hip Exoskeleton

K. Seo

Robotic IntelligenceMixture of ExpertsTime Series

🎯 What it does: Real-time estimation of walking speed and stride length using an embedded IMU in a robotic hip exoskeleton

Real-time event simulation with frame-based cameras

A. Ziegler, A. Zell

Data SynthesisComputational Efficiency

🎯 What it does: Proposed a real-time event simulation method, significantly increasing event simulation speed by an order of magnitude, achieving real-time usability

Real-Time Fast Marching Tree for Mobile Robot Motion Planning in Dynamic Environments

Jefferson Silveira, J. Marshall

OptimizationRobotic Intelligence

🎯 What it does: Propose a Real-Time Fast Marching Tree (RT-FMT) algorithm for path planning of mobile robots in dynamic environments, supporting local and global path generation, reuse of multi-query planning, and dynamic obstacle avoidance.

Real-Time Generative Grasping with Spatio-temporal Sparse Convolution

T. Player, Geoffrey A. Hollinger

GenerationRobotic IntelligenceConvolutional Neural NetworkOptical FlowVideoPoint Cloud

🎯 What it does: Proposed the TSGrasp network, which uses spatiotemporal sparse convolution to process streaming point clouds in real-time and generate 6-DOF grasp poses.

Real-Time Model Predictive Control for Industrial Manipulators with Singularity-Tolerant Hierarchical Task Control

Jaemin Lee, L. Sentis

OptimizationRobotic Intelligence

🎯 What it does: Proposes a real-time model predictive control (MPC) strategy for multi-task control of industrial robotic arms within a finite time horizon. The approach linearizes kinematic and dynamic models using reference trajectories generated by a hierarchical controller, then rapidly solves the linearized MPC problem via quadratic programming, achieving an update frequency exceeding 1 kHz. The method's effectiveness is validated through numerical simulations and experiments on real industrial robotic arms.

Real-Time Navigation for Autonomous Surface Vehicles In Ice-Covered Waters

Rodrigue de Schaetzen, Stephen L. Smith

Autonomous DrivingOptimization

🎯 What it does: Propose a real-time navigation framework aimed at reducing ship-ice collisions and minimizing navigation distance.

Real-Time Reinforcement Learning for Vision-Based Robotics Utilizing Local and Remote Computers

Yan Wang, A. Mahmood

Computational EfficiencyRobotic IntelligenceReinforcement LearningImage

🎯 What it does: Implemented a real-time reinforcement learning system called Remote-Local Distributed (ReLoD), which can distribute the computational tasks of Soft Actor-Critic (SAC) and Proximal Policy Optimization (PPO) between local resource-constrained computers and remote high-performance computers;

Real-Time Simultaneous Localization and Mapping with LiDAR Intensity

Wenqian Du, G. Beltrame

Simultaneous Localization and MappingPoint Cloud

🎯 What it does: Propose a real-time SLAM method based on LiDAR intensity images, which extracts feature points from intensity images for front-end registration by matching them with point clouds, and jointly optimizes feature point distances and plane-point distances in the back-end, while utilizing intensity features for loop closure detection and pose graph optimization.

Real-Time Unified Trajectory Planning and Optimal Control for Urban Autonomous Driving Under Static and Dynamic Obstacle Constraints

Rowan Dempster, W. Melek

Autonomous DrivingOptimization

🎯 What it does: Proposed a unified trajectory planning and control scheme based on Model Predictive Control (MPC) for urban environments, taking into account constraints from static and dynamic obstacles.

Rearrange Indoor Scenes for Human-Robot Co-Activity

Weiqi Wang, Hangxin Liu

OptimizationMeshGraph

🎯 What it does: Proposes an optimization-based framework for rearranging indoor furniture to better support human-robot collaborative activities.

Receding Horizon Planning with Rule Hierarchies for Autonomous Vehicles

Sushant Veer, M. Pavone

Autonomous DrivingOptimizationExplainability and InterpretabilityReinforcement Learning

🎯 What it does: Rule hierarchies are equivalently represented as differentiable reward structures, and a two-stage planner is developed to address conflicting requirements in autonomous vehicle planning.

Reconfigurable Inflated Soft Arms

Nam Gyun Kim, J. Ryu

Robotic Intelligence

🎯 What it does: Designed and experimentally verified a reconfigurable inflatable soft arm structure capable of achieving multiple stable postures under a single inflation.

Reconstructing Objects in-the-wild for Realistic Sensor Simulation

Ze Yang, R. Urtasun

GenerationData SynthesisAutonomous DrivingImagePoint Cloud

🎯 What it does: Propose the NeuSim method, which combines neural signed distance functions with LiDAR and camera data to reconstruct accurate geometry and realistic appearance of objects from sparse views in wild outdoor data, and achieve novel view rendering

Reinforced Learning for Label-Efficient 3D Face Reconstruction

H. Mohaghegh, Bennamoun

RecognitionData-Centric LearningReinforcement LearningImage

🎯 What it does: Propose an active learning framework based on reinforcement learning and clustering-based pooling, actively selecting the most informative views for training a 3D face reconstruction network.

Reinforcement Learning Based Pushing and Grasping Objects from Ungraspable Poses

Hao Zhang, Jian-Wei Zhang

Domain AdaptationRobotic IntelligenceReinforcement LearningAuto EncoderGenerative Adversarial NetworkImage

🎯 What it does: Proposed a model-free deep reinforcement learning framework that jointly plans pushing and grasping actions, and extracts high-dimensional image features using a pre-trained variational autoencoder (VAE); employs a shared-layer Actor-Critic architecture with Proximal Policy Optimization (PPO) algorithm to enable a single network to learn pushing and grasping simultaneously.