arXiv subjects

Pengwei Xie

Publications and source records attributed to Pengwei Xie.

At least 19 recordsLinked to original sources

arXiv2026

$\tau_0$-VLA: a Hierarchical Robot Foundation Model with World-Model-Guided Test-Time Computation

Long-horizon robot manipulation requires a robot to both execute individual skills reliably and sequence them coherently over extended tasks. Most hierarchical vision-language-action (VLA) models make each such decision with a single forward pass, leaving no mechanism to allocate additional computation to difficult or consequential choices. We introduce $\tau_0$-VLA, a hierarchical robot foundation model that formulates high-level subtask generation as a compute-scalable inference problem through world-model-guided test-time computation. At each inference step, the high-level policy uses execution memory to generate a subtask and, when needed, searches over alternatives before committing to its output. A low-level policy then executes the generated subtask across multiple robot embodiments. The policy is trained on 40,115 hours of heterogeneous real-world data with multimodal co-training. Across in-domain and distribution-shifted settings, allocating additional test-time computation substantially improves next-subtask prediction accuracy, and these gains translate into higher closed-loop success on long-horizon robot manipulation tasks.

cs.RO↗

arXiv2026

Dual Modality Prompted Diffusion Priors for Zero Shot Hyperspectral Pansharpening

Hyperspectral pansharpening aims to reconstruct a high resolution hyperspectral (HRHS) image from a panchromatic (PAN) image and a low resolution hyperspectral (LRHS) image while preserving both spatial details and spectral fidelity. Recent diffusion based methods exploit pretrained image priors by generating a low dimensional representation and subsequently mapping it to the hyperspectral domain. However, the observed panchromatic and hyperspectral images are typically imposed only through external reconstruction objectives, limiting their direct interaction with the diffusion prior. To address this issue, we propose dual-modality image-prompted diffusion model (DIDM) for zero shot hyperspectral pansharpening. DIDM encodes the low resolution hyperspectral and panchromatic observations into spectral and spatial prompt tokens, respectively, and injects them into intermediate features of a frozen remote sensing diffusion model through cross attention, allowing complementary spectral and spatial information to directly guide diffusion feature evolution. In addition, we introduce a panchromatic guided weighted pixel aware total variation regularizer that combines low resolution hyperspectral degradation fidelity and panchromatic response fidelity with gradient adaptive structural regularization, thereby preserving structural discontinuities while suppressing spurious variations in homogeneous regions. Extensive experiments on Pavia, Chikusei, and Houston under reduced resolution protocols show that DIDM achieves the best performance across all evaluated metrics, while full resolution evaluation on FR1 yields the highest HQNR among the compared methods. These results demonstrate that internal dual modality prompting and panchromatic guided structural regularization provide an effective balance between spatial detail enhancement and spectral preservation.

cs.CV↗

arXiv2026

Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic Manipulation

Hierarchical Vision-Language-Action (VLA) models have rapidly become a dominant paradigm for robotic manipulation. It typically comprising a Vision-Language backbone for perception and understanding, together with a generative policy for action generation. However, its performance is increasingly bottlenecked by the action generation proceess. (i) Low inference efficiency. A pronounced distributional gap between isotropic noise priors and target action distributions, which increases denoising steps and the incidence of infeasible samples. (ii) Poor robustness. Existing policies condition solely on the current observation, neglecting the constraint of history sequence and thus lacking awareness of task progress and temporal consistency. To address these issues, we introduce OptimusVLA, a dual-memory VLA framework with Global Prior Memory (GPM) and Local Consistency Memory (LCM). GPM replaces Gaussian noise with task-level priors retrieved from semantically similar trajectories, thereby shortening the generative path and reducing the umber of function evaluations (NFE). LCM dynamically models executed action sequence to infer task progress and injects a learned consistency constraint that enforces temporal coherence and smoothness of trajectory. Across three simulation benchmarks, OptimusVLA consistently outperforms strong baselines: it achieves 98.6% average success rate on LIBERO, improves over pi_0 by 13.5% on CALVIN, and attains 38% average success rate on RoboTwin 2.0 Hard. In Real-World evaluation, OptimusVLA ranks best on Generalization and Long-horizon suites, surpassing pi_0 by 42.9% and 52.4%, respectively, while delivering 2.9x inference speedup.

cs.RO↗

arXiv2026

Learning While Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies

Generalist robot policies increasingly benefit from large-scale pretraining, but offline data alone is insufficient for robust real-world deployment. Deployed robots encounter distribution shifts, long-tail failures, task variations, and human correction opportunities that fixed demonstration datasets cannot fully capture. We present Learning While Deploying (LWD), a fleet-scale offline-to-online reinforcement learning framework for continual post-training of generalist Vision-Language-Action (VLA) policies. Starting from a pretrained VLA policy, LWD closes the loop between deployment, shared physical experience, policy improvement, and redeployment by using autonomous rollouts and human interventions collected across a robot fleet. To stabilize learning from heterogeneous, sparse-reward fleet data, LWD combines Distributional Implicit Value Learning (DIVL) for robust value estimation with Q-learning via Adjoint Matching (QAM) for policy extraction in flow-based VLA action generators. We validate LWD on a fleet of 16 dual-arm robots across eight real-world manipulation tasks, including semantic grocery restocking and 3--5 minute long-horizon tasks. A single generalist policy improves as fleet experience accumulates, reaching an average success rate of 95%, with the largest gains on long-horizon tasks.

cs.RO↗

arXiv2026

OmniEVA: Embodied Versatile Planner via Task-Adaptive 3D-Grounded and Embodiment-aware Reasoning

Recent advances in multimodal large language models (MLLMs) have opened new opportunities for embodied intelligence, enabling multimodal understanding, reasoning, and interaction, as well as continuous spatial decision-making. Nevertheless, current MLLM-based embodied systems face two critical limitations. First, Geometric Adaptability Gap: models trained solely on 2D inputs or with hard-coded 3D geometry injection suffer from either insufficient spatial information or restricted 2D generalization, leading to poor adaptability across tasks with diverse spatial demands. Second, Embodiment Constraint Gap: prior work often neglects the physical constraints and capacities of real robots, resulting in task plans that are theoretically valid but practically infeasible. To address these gaps, we introduce OmniEVA -- an embodied versatile planner that enables advanced embodied reasoning and task planning through two pivotal innovations: (1) a Task-Adaptive 3D Grounding mechanism, which introduces a gated router to perform explicit selective regulation of 3D fusion based on contextual requirements, enabling context-aware 3D grounding for diverse embodied tasks. (2) an Embodiment-Aware Reasoning framework that jointly incorporates task goals and embodiment constraints into the reasoning loop, resulting in planning decisions that are both goal-directed and executable. Extensive experimental results demonstrate that OmniEVA not only achieves state-of-the-art general embodied reasoning performance, but also exhibits a strong ability across a wide range of downstream scenarios. Evaluations of a suite of proposed embodied benchmarks, including both primitive and composite tasks, confirm its robust and versatile planning capabilities. Project page: https://omnieva.github.io

cs.RO↗

arXiv2025

More than A Point: Capturing Uncertainty with Adaptive Affordance Heatmaps for Spatial Grounding in Robotic Tasks

Many language-guided robotic systems rely on collapsing spatial reasoning into discrete points, making them brittle to perceptual noise and semantic ambiguity. To address this challenge, we propose RoboMAP, a framework that represents spatial targets as continuous, adaptive affordance heatmaps. This dense representation captures the uncertainty in spatial grounding and provides richer information for downstream policies, thereby significantly enhancing task success and interpretability. RoboMAP surpasses the previous state-of-the-art on a majority of grounding benchmarks with up to a 50x speed improvement, and achieves an 82\% success rate in real-world manipulation. Across extensive simulated and physical experiments, it demonstrates robust performance and shows strong zero-shot generalization to navigation. More details and videos can be found at https://robo-map.github.io.

cs.RO↗

arXiv2024

Region-aware Grasp Framework with Normalized Grasp Space for Efficient 6-DoF Grasping

A series of region-based methods succeed in extracting regional features and enhancing grasp detection quality. However, faced with a cluttered scene with potential collision, the definition of the grasp-relevant region stays inconsistent, and the relationship between grasps and regional spaces remains incompletely investigated. In this paper, we propose Normalized Grasp Space (NGS) from a novel region-aware viewpoint, unifying the grasp representation within a normalized regional space and benefiting the generalizability of methods. Leveraging the NGS, we find that CNNs are underestimated for 3D feature extraction and 6-DoF grasp detection in clutter scenes and build a highly efficient Region-aware Normalized Grasp Network (RNGNet). Experiments on the public benchmark show that our method achieves significant >20% performance gains while attaining a real-time inference speed of approximately 50 FPS. Real-world cluttered scene clearance experiments underscore the effectiveness of our method. Further, human-to-robot handover and dynamic object grasping experiments demonstrate the potential of our proposed method for closed-loop grasping in dynamic scenarios.

cs.RO↗

arXiv2024

Rethinking 6-Dof Grasp Detection: A Flexible Framework for High-Quality Grasping

Robotic grasping is a primitive skill for complex tasks and is fundamental to intelligence. For general 6-Dof grasping, most previous methods directly extract scene-level semantic or geometric information, while few of them consider the suitability for various downstream applications, such as target-oriented grasping. Addressing this issue, we rethink 6-Dof grasp detection from a grasp-centric view and propose a versatile grasp framework capable of handling both scene-level and target-oriented grasping. Our framework, FlexLoG, is composed of a Flexible Guidance Module and a Local Grasp Model. Specifically, the Flexible Guidance Module is compatible with both global (e.g., grasp heatmap) and local (e.g., visual grounding) guidance, enabling the generation of high-quality grasps across various tasks. The Local Grasp Model focuses on object-agnostic regional points and predicts grasps locally and intently. Experiment results reveal that our framework achieves over 18% and 23% improvement on unseen splits of the GraspNet-1Billion Dataset. Furthermore, real-world robotic tests in three distinct settings yield a 95% success rate.

cs.RO↗

arXiv2024

GAP-RL: Grasps As Points for RL Towards Dynamic Object Grasping

Dynamic grasping of moving objects in complex, continuous motion scenarios remains challenging. Reinforcement Learning (RL) has been applied in various robotic manipulation tasks, benefiting from its closed-loop property. However, existing RL-based methods do not fully explore the potential for enhancing visual representations. In this letter, we propose a novel framework called Grasps As Points for RL (GAP-RL) to effectively and reliably grasp moving objects. By implementing a fast region-based grasp detector, we build a Grasp Encoder by transforming 6D grasp poses into Gaussian points and extracting grasp features as a higher-level abstraction than the original object point features. Additionally, we develop a Graspable Region Explorer for real-world deployment, which searches for consistent graspable regions, enabling smoother grasp generation and stable policy execution. To assess the performance fairly, we construct a simulated dynamic grasping benchmark involving objects with various complex motions. Experiment results demonstrate that our method effectively generalizes to novel objects and unseen dynamic motions compared to other baselines. Real-world experiments further validate the framework's sim-to-real transferability.

cs.RO↗

arXiv2024

Constraints on the axial-vector and pseudo-scalar mediated WIMP-nucleus interactions from PandaX-4T experiment

We present the constraints on the axial-vector and pseudo-scalar mediated WIMP-nucleus interactions from the PandaX-4T experiment, using the data set corresponding to a total exposure of 0.63~tonne$\cdot$year. No significant signal excess is observed, and the most stringent upper limits to date on the spin-dependent WIMP-neutron scattering cross section are set at 90\% confidence level with the minimum WIMP-neutron scattering cross section of 5.8$\times 10^{-42}$\si{\cm^{2}} for WIMP mass of 40~\si{\GeV/}$c^2$. Exclusion limits on the axial-vector and pseudo-scalar simplified models are also derived.

hep-ex↗

arXiv2024

Target-Oriented Object Grasping via Multimodal Human Guidance

In the context of human-robot interaction and collaboration scenarios, robotic grasping still encounters numerous challenges. Traditional grasp detection methods generally analyze the entire scene to predict grasps, leading to redundancy and inefficiency. In this work, we reconsider 6-DoF grasp detection from a target-referenced perspective and propose a Target-Oriented Grasp Network (TOGNet). TOGNet specifically targets local, object-agnostic region patches to predict grasps more efficiently. It integrates seamlessly with multimodal human guidance, including language instructions, pointing gestures, and interactive clicks. Thus our system comprises two primary functional modules: a guidance module that identifies the target object in 3D space and TOGNet, which detects region-focal 6-DoF grasps around the target, facilitating subsequent motion planning. Through 50 target-grasping simulation experiments in cluttered scenes, our system achieves a success rate improvement of about 13.7%. In real-world experiments, we demonstrate that our method excels in various target-oriented grasping scenarios.

cs.RO↗

arXiv2024

Signal Response Model in PandaX-4T

PandaX-4T experiment is a deep-underground dark matter direct search experiment that employs a dual-phase time projection chamber with a sensitive volume containing 3.7 tonne of liquid xenon. The detector of PandaX-4T is capable of simultaneously collecting the primary scintillation and ionization signals, utilizing their ratio to discriminate dark matter signals from background sources such as gamma rays and beta particles. The signal response model plays a crucial role in interpreting the data obtained by PandaX-4T. It describes the conversion from the deposited energy by dark matter interactions to the detectable signals within the detector. The signal response model is utilized in various PandaX-4T results. This work provides a comprehensive description of the procedures involved in constructing and parameter-fitting the signal response model for the energy range of approximately 1 keV to 25 keV for electronic recoils and 6 keV to 90 keV for nuclear recoils. It also covers the signal reconstruction, selection, and correction methods, which are crucial components integrated into the signal response model.

physics.ins-det↗

arXiv2024

GenH2R: Learning Generalizable Human-to-Robot Handover via Scalable Simulation, Demonstration, and Imitation

This paper presents GenH2R, a framework for learning generalizable vision-based human-to-robot (H2R) handover skills. The goal is to equip robots with the ability to reliably receive objects with unseen geometry handed over by humans in various complex trajectories. We acquire such generalizability by learning H2R handover at scale with a comprehensive solution including procedural simulation assets creation, automated demonstration generation, and effective imitation learning. We leverage large-scale 3D model repositories, dexterous grasp generation methods, and curve-based 3D animation to create an H2R handover simulation environment named \simabbns, surpassing the number of scenes in existing simulators by three orders of magnitude. We further introduce a distillation-friendly demonstration generation method that automatically generates a million high-quality demonstrations suitable for learning. Finally, we present a 4D imitation learning method augmented by a future forecasting objective to distill demonstrations into a visuo-motor handover policy. Experimental evaluations in both simulators and the real world demonstrate significant improvements (at least +10\% success rate) over baselines in all cases. The project page is https://GenH2R.github.io/.

cs.RO↗

arXiv2024

Waveform Simulation in PandaX-4T

Signal reconstruction through software processing is a crucial component of the background and signal models in the PandaX-4T experiment, which is a multi-tonne dark matter direct search experiment. The accuracy of signal reconstruction is influenced by various detector artifacts, including noise, dark count of photomultiplier, impurity photoionization in the detector, and other relevant considerations. In this study, we present a detailed description of a semi-data-driven approach designed to simulate the signal waveform. This work provides a reliable model for the efficiency and bias of the signal reconstruction in the data analysis of PandaX-4T. By comparing critical variables which relate to the temporal shape and hit pattern of the signals, we demonstrate a good agreement between the simulation and data.

hep-ex↗

arXiv2024

Efficient Heatmap-Guided 6-Dof Grasp Detection in Cluttered Scenes

Fast and robust object grasping in clutter is a crucial component of robotics. Most current works resort to the whole observed point cloud for 6-Dof grasp generation, ignoring the guidance information excavated from global semantics, thus limiting high-quality grasp generation and real-time performance. In this work, we show that the widely used heatmaps are underestimated in the efficiency of 6-Dof grasp generation. Therefore, we propose an effective local grasp generator combined with grasp heatmaps as guidance, which infers in a global-to-local semantic-to-point way. Specifically, Gaussian encoding and the grid-based strategy are applied to predict grasp heatmaps as guidance to aggregate local points into graspable regions and provide global semantic information. Further, a novel non-uniform anchor sampling mechanism is designed to improve grasp accuracy and diversity. Benefiting from the high-efficiency encoding in the image space and focusing on points in local graspable regions, our framework can perform high-quality grasp detection in real-time and achieve state-of-the-art results. In addition, real robot experiments demonstrate the effectiveness of our method with a success rate of 94% and a clutter completion rate of 100%. Our code is available at https://github.com/THU-VCLab/HGGD.

cs.RO↗

arXiv2024

Searching for Two-Neutrino and Neutrinoless Double Beta Decay of $^{134}$Xe with the PandaX-4T Experiment

$^{134}$Xe is a candidate isotope for neutrinoless double beta decay~($0νββ$) search. In addition, the two-neutrino case ($2νββ$) allowed by the Standard Model of particle physics has not yet been observed. Utilizing the 10.4% of $^{134}$Xe in the natural xenon in the PandaX-4T detector and its first 94.9-day exposure, we have established the most stringent constraints on $2νββ$ and $0νββ$ of $^{134}$Xe half-lives, with limits of $2.8\times10^{22}$ yr and $3.0\times10^{23}$ yr at 90% confidence level, respectively. The $2νββ$ ($0νββ$) limit surpasses the previously reported best result by a factor of 32 (2.7), highlighting the potential of large monolithic natural xenon detectors.

nucl-ex↗

arXiv2024

Part-Guided 3D RL for Sim2Real Articulated Object Manipulation

Manipulating unseen articulated objects through visual feedback is a critical but challenging task for real robots. Existing learning-based solutions mainly focus on visual affordance learning or other pre-trained visual models to guide manipulation policies, which face challenges for novel instances in real-world scenarios. In this paper, we propose a novel part-guided 3D RL framework, which can learn to manipulate articulated objects without demonstrations. We combine the strengths of 2D segmentation and 3D RL to improve the efficiency of RL policy training. To improve the stability of the policy on real robots, we design a Frame-consistent Uncertainty-aware Sampling (FUS) strategy to get a condensed and hierarchical 3D representation. In addition, a single versatile RL policy can be trained on multiple articulated object manipulation tasks simultaneously in simulation and shows great generalizability to novel categories and instances. Experimental results demonstrate the effectiveness of our framework in both simulation and real-world settings. Our code is available at https://github.com/THU-VCLab/Part-Guided-3D-RL-for-Sim2Real-Articulated-Object-Manipulation.

cs.RO↗

arXiv2024

Category-Agnostic Pose Estimation for Point Clouds

The goal of object pose estimation is to visually determine the pose of a specific object in the RGB-D input. Unfortunately, when faced with new categories, both instance-based and category-based methods are unable to deal with unseen objects of unseen categories, which is a challenge for pose estimation. To address this issue, this paper proposes a method to introduce geometric features for pose estimation of point clouds without requiring category information. The method is based only on the patch feature of the point cloud, a geometric feature with rotation invariance. After training without category information, our method achieves as good results as other category-based methods. Our method successfully achieved pose annotation of no category information instances on the CAMERA25 dataset and ModelNet40 dataset.

cs.CV↗