All papers

This paper presents a training-free method to adaptively select action chunks and execution horizons to speed up robot policy inference.

This paper presents an orchestration layer that detects failures and backtracks along executed trajectories to re-attempt robot tasks.

This paper proposes a framework enabling vision-language-action models to adapt to novel manipulation tasks using a single human video demonstration.

This paper introduces an interleaved dual-hand action representation to improve cross-hand coordination in bimanual dexterous manipulation models.

This paper proposes a feedback control framework for combining control inputs and feedback signals across simulation and physical setups.

This paper introduces a method that predicts action errors to decide when and how much to refine robot action proposals.

This paper presents a modular actuation platform designed to standardize hardware and software across different continuum robot designs.

This paper presents a symbolic supervisory control architecture to formally certify safety and reachability for learned robot motion models.

This paper introduces a handheld interface that transmits physical contact forces to human operators without using force sensors.

This paper proposes an active perception method for robots to estimate human upper-limb functional mobility through repeated handover interactions.

This paper introduces a training-free token pruning method to accelerate inference in world action models without degrading action performance.

This paper introduces an indoor navigation benchmark for teams of robot agents searching for multiple multi-modal object goals.

This paper surveys learning-based tactile manipulation, covering representations, policy learning, foundation models, and sim-to-real transfer.

This paper presents a self-locking deployable linkage that enables robotic vacuum cleaners to support loads without continuous motor torque.

This paper introduces a framework that connects task reasoning to robotic action generation via intermediate atomic-motion descriptions.

The paper presents an object-centric vision-language-action framework that tracks persistent object slots and predicts future states for manipulation.

The paper proposes a visual quality estimator to adaptively balance camera and IMU weighting during bundle adjustment.

The paper introduces a hierarchical navigation policy that determines candidate routes from local geometry to prevent averaging conflicting demonstrations.

The paper introduces polynomial-time approximation schemes for multi-robot motion planning with polygonal obstacles and robots.

The paper proposes a receding-horizon safety critic and differentiable shield to prevent robot failures in reinforcement learning.

The paper introduces an adaptive model predictive control algorithm for CubeSat attitude control using only magnetorquers.

The paper introduces a large multimodal dataset of paired visual, tactile, and trajectory demonstrations for dexterous robot manipulation.

The paper proposes a student-teacher framework that distills long-horizon model predictive control into vision-guided short-horizon control.

The paper presents an automated pipeline that converts real workspace scans into simulation assets for scalable robot learning.

The paper compares discrete and continuous action parametrizations across robot foundation models in simulation and physical tests.

The paper provides a survey and unifying taxonomy of foundation models and systems for agentic robots.

The paper proposes a distributed dynamic belief propagation method for real-time trajectory planning across multiple robot arms.

The paper introduces a tiered-resolution mapping method that integrates high-resolution LiDAR data only along planned paths.

The paper establishes a dual certification framework for non-convex robot state estimation under severe sensor degeneracy.

The paper presents a graph-based reweighting method to train manipulation policies from mixed-quality demonstrations and intervention data.

The paper incorporates continuous physical time into robot policy memory to improve action timing across variable observation intervals.

The paper introduces an open-source tool enabling robots to act on spoken instructions before the speaker finishes talking.

The paper uses kinematically grounded video generation to synthesize customizable reference motions for humanoid robot loco-manipulation.

The paper presents a coordination framework for humanoid robots playing table tennis doubles with collision-avoidance planning.

The paper introduces a multi-device dataset and collaborative framework for wearable inertial human motion localization.

The paper introduces a benchmark and model for autonomous driving vision-language-action systems handling conflicting or unsafe navigation instructions.

The paper characterizes scan-induced artifacts in dynamic 3D meshes to build a no-reference visual quality assessment method.

The paper improves behavior cloning for autonomous driving using bird's-eye view inputs and action-density weighting for rare events like traffic lights.

This paper presents an entropy-regularized memory-guided policy optimization method for online test-time adaptation in robot locomotion.

HumanVerse provides an egocentric dataset and alignment pipeline to transfer human loco-manipulation behaviors directly to humanoid robots without target-task fine-tuning.

Action Jets models the local shape of action-conditioned manifolds via Taylor approximations to enable efficient short-horizon robotic planning.

The study evaluates coding agents generating images via executable code, analyzing their strengths in geometric constraints over photorealistic drawing.

RingNav is a compact visual heading estimator that restricts persistent memory updates to a circular manifold to improve trajectory tracking.

CardioFAD resolves temporal ambiguity in sparse cardiac dynamic synthesis by using ECG-aligned motion-salient keyframe diffusion.

VITAL introduces visual-physiological token alignment to enable vision-language models to perform remote physiological measurement alongside general visual reasoning.

WorldMemBench introduces an evaluation benchmark assessing static entity retention and dynamic event persistence in video world models over multi-revisit paths.

MonoMIP combines multi-height inverse perspective mapping with a dual-decoder framework to improve bird's-eye view reasoning in monocular 3D object detection.

A meta-analysis shows multimodal language models struggle with geometric acquisition variations in aerial video, exposing limitations in standard QA-based evaluations.

DoReMe uses Gaussian splatting primitives to jointly optimize image reconstruction and respiratory motion estimation in self-navigated free-breathing abdominal MRI.

The study evaluates field pea breeding in Iowa by characterizing agronomic yield performance and extracting time-series traits from drone phenotyping data.

LidarVerse provides a unified conditional generative LiDAR world model combining geometry-preserving spatial caches with amortized sequence learning across sensors and tasks.

D3WM decouples deterministic state transitions from spatial uncertainty to guide conditional diffusion for consistent future frame prediction in world models.

A gradient-regularized U-Net framework transforms images to counteract spherical eye defocus aberrations and improve visual clarity for myopic individuals.

The paper investigates the emergent dynamics of the Continuous Thought Machine during image classification using non-normal maps and fixed-point curves.

SMART introduces a slot-memory mechanism addressed by region tokens to achieve bounded memory and spatially consistent recall in long-horizon video generation.

The paper proposes a framework using vision-language reasoning to estimate point-conditioned 3D directions and physical light source contributions.

The paper compares GAN and Transformer architectures for mitigating atmospheric turbulence distortions in multi-frame imaging.

The paper presents open speech-to-video models that generate long-form avatar video with expressive motion driven by an audio waveform.

The paper proposes a data-free multistep distillation method for text-to-image diffusion models to improve generation fidelity across sampling budgets.

The paper proposes an interactive ultrasound world model that learns video dynamics and maps them to six-degree-of-freedom probe control.