arXiv subjects

Qiyao Zhang

Publications and source records attributed to Qiyao Zhang.

8 recordsLinked to original sources

arXiv2026

AIR-VLA+: Decoupling Movement and Manipulation via Cascaded Dual-Action Decoders with Asymmetric MoE for Aerial Robots

Aerial manipulation systems have long suffered from representation coupling in end-to-end control, as platform-level Unmanned Aerial Vehicle (UAV) movement and end-effector-level arm manipulation differ substantially in action scale, dynamics, and control objectives. In this paper, we propose AIR-VLA+, a flow matching action generation architecture specifically designed for aerial manipulation, featuring cascaded dual-action decoders and an asymmetric feature-level Mixture of Experts (MoE). We construct cascaded manipulation and movement decoders, allowing the UAV to unidirectionally observe the manipulator's intent during movement to achieve workflow coordination, while isolating the impact of UAV movement information backpropagation on arm manipulation stability. Addressing the characteristic that UAV movement is highly dependent on high-level semantics and responsible for task state transitions in aerial manipulation, we design an input feature enhancement module for the UAV movement decoder. This module introduces an implicit visual grasp projector to perceive the interaction state between the gripper and the object, and injects compressed global semantic features. Within the UAV movement decoder, we deploy an implicit MoE architecture, enabling different movement experts to spontaneously exhibit capacity inclinations for various task stages during training. Through dense soft blending computation on the feature manifold, the UAV movement is endowed with stronger task-stage adaptability. Experiments on the standardized AIR-VLA benchmark demonstrate that our method comprehensively surpasses all baselines with an overall average score of 48.0. The overall task completion score improves by 80.2\% compared to the single-head $\pi_{0.5}$ policy, effectively mitigating the heterogeneous coordinated control conflicts of composite robots.

cs.RO↗

arXiv2026

UAV-Track VLA: Embodied Aerial Tracking via Vision-Language-Action Models

Embodied visual tracking is crucial for Unmanned Aerial Vehicles (UAVs) executing complex real-world tasks. In dynamic urban scenarios with complex semantic requirements, Vision-Language-Action (VLA) models show great promise due to their cross-modal fusion and continuous action generation capabilities. To benchmark multimodal tracking in such environments, we construct a dedicated evaluation benchmark and a large-scale dataset encompassing over 890K frames, 176 tasks, and 85 diverse objects. Furthermore, to address temporal feature redundancy and the lack of spatial geometric priors in existing VLA models, we propose an improved VLA tracking model, UAV-Track VLA. Built upon the $π_{0.5}$ architecture, our model introduces a temporal compression net to efficiently capture inter-frame dynamics. Additionally, a parallel dual-branch decoder comprising a spatial-aware auxiliary grounding head and a flow matching action expert is designed to decouple cross-modal features and generate fine-grained continuous actions. Systematic experiments in the CARLA simulator validate the superior end-to-end performance of our method. Notably, in challenging long-distance pedestrian tracking tasks, UAV-Track VLA achieves a 61.76\% success rate and 269.65 average tracking frames, significantly outperforming existing baselines. Furthermore, it demonstrates robust zero-shot generalization in unseen environments and reduces single-step inference latency by 33.4\% (to 0.0571s) compared to the original $π_{0.5}$, enabling highly efficient, real-time UAV control. Data samples and demonstration videos are available at: https://github.com/Hub-Tian/UAV-Track_VLA.

cs.CV↗

arXiv2026

AIR-VLA: Vision-Language-Action Systems for Aerial Manipulation

While Vision-Language-Action (VLA) models have achieved remarkable success in ground-based embodied intelligence, their application to Aerial Manipulation Systems (AMS) remains a largely unexplored frontier. The inherent characteristics of AMS, including floating-base dynamics, strong coupling between the UAV and the manipulator, and the multi-step, long-horizon nature of operational tasks, pose severe challenges to existing VLA paradigms designed for static or 2D mobile bases. To bridge this gap, we propose \textbf{AIR-VLA}, the first VLA benchmark specifically tailored for aerial manipulation. We construct a physics-based simulation environment and release a high-quality multimodal dataset comprising 3000 manually teleoperated demonstrations, covering base manipulation, object \& spatial understanding, semantic reasoning, and long-horizon planning. Leveraging this platform, we systematically evaluate mainstream VLA models and state-of-the-art VLM models. Our experiments not only validate the feasibility of transferring VLA paradigms to aerial systems but also, through multi-dimensional metrics tailored to aerial tasks, reveal the capabilities and boundaries of current models regarding UAV mobility, manipulator control, and high-level planning. \textbf{AIR-VLA} establishes a standardized testbed and data foundation for future research in general-purpose aerial robotics. The resource of AIR-VLA will be available at https://github.com/SpencerSon2001/AIR-VLA.

cs.RO↗

arXiv2026

CogRail: Benchmarking VLMs in Cognitive Intrusion Perception for Intelligent Railway Transportation Systems

Accurate and early perception of potential intrusion targets is essential for ensuring the safety of railway transportation systems. However, most existing systems focus narrowly on object classification within fixed visual scopes and apply rule-based heuristics to determine intrusion status, often overlooking targets that pose latent intrusion risks. Anticipating such risks requires the cognition of spatial context and temporal dynamics for the object of interest (OOI), which presents challenges for conventional visual models. To facilitate deep intrusion perception, we introduce a novel benchmark, CogRail, which integrates curated open-source datasets with cognitively driven question-answer annotations to support spatio-temporal reasoning and prediction. Building upon this benchmark, we conduct a systematic evaluation of state-of-the-art visual-language models (VLMs) using multimodal prompts to identify their strengths and limitations in this domain. Furthermore, we fine-tune VLMs for better performance and propose a joint fine-tuning framework that integrates three core tasks, position perception, movement prediction, and threat analysis, facilitating effective adaptation of general-purpose foundation models into specialized models tailored for cognitive intrusion perception. Extensive experiments reveal that current large-scale multimodal models struggle with the complex spatial-temporal reasoning required by the cognitive intrusion perception task, underscoring the limitations of existing foundation models in this safety-critical domain. In contrast, our proposed joint fine-tuning framework significantly enhances model performance by enabling targeted adaptation to domain-specific reasoning demands, highlighting the advantages of structured multi-task learning in improving both accuracy and interpretability. Code will be available at https://github.com/Hub-Tian/CogRail.

cs.CV↗

arXiv2025

UAVs Meet LLMs: Overviews and Perspectives Toward Agentic Low-Altitude Mobility

Low-altitude mobility, exemplified by unmanned aerial vehicles (UAVs), has introduced transformative advancements across various domains, like transportation, logistics, and agriculture. Leveraging flexible perspectives and rapid maneuverability, UAVs extend traditional systems' perception and action capabilities, garnering widespread attention from academia and industry. However, current UAV operations primarily depend on human control, with only limited autonomy in simple scenarios, and lack the intelligence and adaptability needed for more complex environments and tasks. The emergence of large language models (LLMs) demonstrates remarkable problem-solving and generalization capabilities, offering a promising pathway for advancing UAV intelligence. This paper explores the integration of LLMs and UAVs, beginning with an overview of UAV systems' fundamental components and functionalities, followed by an overview of the state-of-the-art in LLM technology. Subsequently, it systematically highlights the multimodal data resources available for UAVs, which provide critical support for training and evaluation. Furthermore, it categorizes and analyzes key tasks and application scenarios where UAVs and LLMs converge. Finally, a reference roadmap towards agentic UAVs is proposed, aiming to enable UAVs to achieve agentic intelligence through autonomous perception, memory, reasoning, and tool utilization. Related resources are available at https://github.com/Hub-Tian/UAVs_Meet_LLMs.

cs.RO↗

arXiv2024

The Quadruplon in a Monolayer Semiconductor

Understanding the structure of matter or materials and interaction or correlations among the constituent elementary particles are the central tasks of all branches of science, from physics, chemistry, to biology. In physics, this ultimate goal has spurred a constant search for high-order correlated entities or composite particles for nearly all states and forms of matter, from elementary particles, nuclei, cold atoms, to condensed matter. So far, such composite particles involving two or three constituent particles have been experimentally identified, such as the Cooper pairs, excitons, and trions in condensed matter physics, or diquarks and mesons in quantum chromodynamics. Although the four-body irreducible entities have long been predicted theoretically in a variety of materials systems alternatively as quadruplons, quadrons, or quartets, the closely related experimental observation so far seems to be restricted to the field of elementary particles (e.g. the recent tetraquark at CERN) only. In this article, we present the first experimental evidence for the existence of a four-body irreducible entity, the quadruplon, involving two electrons and two holes in a monolayer of Molybdenum Ditelluride. Using the optical pump-probe technique, we discovered a series of new spectral features that are distinct from those of trions and bi-excitons. By solving the four-body Bethe-Salpeter equation in conjunction with the cluster expansion approach, we are able to explain these spectral features in terms of the four-body irreducible cluster or the quadruplons. In contrast to a bi-exciton which consists of two weakly bound excitons, a quadruplon consists of two electrons and two holes without the presence of an exciton.

cond-mat.mtrl-sci↗

arXiv2021

Prolonging Valley Polarization Lifetime through Gate-Controlled Exciton-to-Trion Conversion in Monolayer Molybdenum Ditelluride

Monolayer 2D semiconductors provide an attractive option for valleytronics due to the valley-addressability by helicity-specific light beam. But the short valley lifetime for excitons have hindered potential valleytronic applications. In this paper, we demonstrate a strategy for prolonging the valley lifetime by converting excitons to trions through effective gate control and by taking advantage of much longer valley lifetime for trions than for excitons. In continuous-wave experiments, we found the valley polarization increases as gate voltage is tuned away from the charge neutrality, with the degree of valley polarization increased from near zero to 38 % for excitons and to 33 % for trions. This is the first successful observation of valley-polarization in MoTe2 without a magnetic field. In pump-probe experiments, we found that the intervalley scattering process of excitons is significantly suppressed as gate voltage is tuned away from charge neutrality, with scattering time from 0.85 ps to ~ 2.17 ps. In contrast, the intervalley scattering rate for trions increases due to increased availability of partner charges for trion spin flipping, with scattering time from 1.39 ns down to ~100 ps away from charge neutrality. Interestingly, our results show that, despite the accelerated intervalley scattering, the trion polarization degree increases due to polarized trion generation from the exciton-to-trion conversion overtaking the intervalley trion scatterings. Importantly, the efficient exciton-to-trion conversion changed the dominant depolarization mechanisms. As a result, the valley lifetime is dramatically improved by 1000 times from excitons to trions at the charge neutrality. Our results shed new light into the depolarization dynamics and the interplay of various depolarization channels for excitons and trions and provide an effective strategy for prolonging the valley polarization.

cond-mat.mes-hall↗

arXiv2018

Excitonic Complexes and Optical Gain in Two-Dimensional Molybdenum Ditelluride Well below Mott Transition

Strong Coulomb interaction in 2D materials provides unprecedented opportunities for studying many key issues of condensed matter physics, such as co-existence and mutual conversions of excitonic complexes, fundamental optical processes associated with their conversions, and their roles in the celebrated Mott transition. Recent lasing demonstrations in 2D materials raise important questions about the existence and origin of optical gain and possible roles of excitonic complexes. While lasing occurred at extremely low densities dominated by various excitonic complexes, optical gain was observed in the only experiment at densities several orders of magnitude higher, exceeding the Mott density. Here, we report a new gain mechanism involving charged excitons or trions well below the Mott density in 2D molybdenum ditelluride. Our combined experimental and modeling study not only reveals the complex interplays of excitonic complexes well below the Mott transition, but also provides foundation for lasing at extremely low excitation levels, important for future energy efficient photonic devices.

cond-mat.mes-hall↗