arXiv subjects

Helong Huang

Publications and source records attributed to Helong Huang.

9 recordsLinked to original sources

arXiv2026

Whole-Body Inverse Kinematics with Graph Diffusion

Inverse kinematics (IK) is a fundamental problem in robotics, requiring the generation of joint configurations that satisfy target end-effector poses. Existing approaches often struggle to generalize across diverse robot morphologies and to effectively model the multi-modal nature of IK, particularly in articulated systems with multiple kinematic branches. In this work, we propose GraphDiff-IK, a structure-aware graph diffusion framework for inverse kinematics. Specifically, we represent the robot as a kinematic graph constructed from the robot URDF, where nodes correspond to actuated joints and edges encode kinematic dependencies. Building upon this representation, we formulate IK as a conditional graph diffusion process that directly generates joint configurations on the robot graph. To better capture structural dependencies in articulated systems, we further introduce a structure-aware graph reasoning framework with hierarchical stage-wise message passing and torso-aware conditioning for multi-branch robots. In addition, we incorporate noisy forward kinematics feedback and task-space supervision to improve geometric consistency during denoising. The proposed framework provides a unified formulation that naturally supports single-arm robots, dual-arm systems, and articulated robots with torso or waist structures. Extensive experiments on diverse robotic platforms demonstrate that the proposed method achieves accurate and stable IK performance while preserving the ability to generate multiple feasible solutions for redundant robotic systems.

cs.RO↗

arXiv2026

ForceFlow: Learning to Feel and Act via Contact-Driven Flow Matching

Existing imitation learning methods enable robots to interact autonomously with the physical environment. However, contact-rich manipulation tasks remain a significant challenge due to complex contact dynamics that demand high-precision force feedback and control. Although recent efforts have attempted to integrate force/torque sensing into policies, how to build a simple yet effective framework that achieves robust generalization under multimodal observations remains an open question. In this paper, we propose ForceFlow, a force-aware reactive framework built upon flow matching. For contact-stage policy design, we investigate force signal fusion mechanisms and adopt an asymmetric multimodal fusion architecture that treats force as a global regulatory signal, combined with a joint prediction paradigm that enhances the policy's understanding of instantaneous force and historical information, thereby achieving deep coupling between force and motion. For task-level hierarchical decomposition, we divide manipulation into a vision-dominant approach stage (VLM-based pointing for target localization) and a touch-dominant interaction stage (force-driven contact execution), with a Vision-to-Force (V2F) handover mechanism that explicitly decouples spatial generalization from contact regulation. Experimental results across six real-world contact-rich tasks demonstrate that ForceFlow achieves a 37% success rate improvement over the strong baseline ForceVLA while maintaining significantly lower cost. Moreover, ForceFlow exhibits accurate force signal prediction and demonstrates superior performance in contact force self-regulation and zero-shot out-of-distribution (OOD) generalization.

cs.RO↗

arXiv2026

A Cloudy Fit to the Atmosphere of WASP-107 b

Context. WASP-107 b has been observed comprehensively by JWST in the near- and mid-IR bands, making it an ideal planet to probe the composition and internal dynamics. Recent analysis reveals a 8-10 um silicate feature, but it still remains uncertain how silicate clouds form on this planet. Aims. We aim at fitting the complete JWST spectrum of WASP-107 b, from 0.9 um to 12 um with a physically motivated cloud model and self-consistent temperature profile. Methods. Two-stream radiative transfer is coupled to a cloud formation model until convergence between cloud and temperature profiles is reached. We search a model grid spanning metallicity, turbulent diffusivity, internal heat flux and nucleation parameters to find the best fit model. Results. The silicate cloud feature at 10 um and the near-IR molecular band strength can be simultaneously and naturally explained without assuming a parametrized temperature profile. A moderate vertical diffusivity of Kzz = 10^9 cm^2 s^-1 is needed to bring the cloud particles to the upper atmosphere of WASP-107 b. This Kzz is favored by the joint fitting of the near-IR water feature and mid-IR silicate feature -- both sensitive to clouds. From the strength of H2O and CO2 bands, our model suggests a metallicity 17 times solar. Conclusions. Even in warm planets such as WASP-107 b, silicate clouds can form in the relatively cool upper atmosphere because turbulence uplifts vapor and cloud particles. Despite having considerably fewer degrees of freedom, the self-consistent modeling approach successfully fits WASP-107 b's multi-wavelength data, instilling confidence in the derived physical parameters.

astro-ph.EP↗

arXiv2026

Hot and cloudy: High temperature clouds in super-Earths and sub-Neptunes

JWST observations provide for the first time evidence for an atmosphere on a rocky exoplanet - 55 Cnc e. The atmosphere of 55 Cnc e is hot with $\text{T}_{\text{eq}}>2000$K and shows strong variability, for which cloud formation above a molten crust could be one possible explanation. The composition of the atmosphere of 55 Cnc e is still unknown but suggests the presence of volatiles. We have run cloud formation models on a grid of N-dominated, O-dominated, C-dominated and H-dominated atmospheres to investigate which type of cloud we could expect on hot super-Earths and hot sub-Neptunes ($1000$K $<$ T $<$ $3000$K). Our models combine radiative transfer with equilibrium chemistry of the gaseous and condensed phases, vertical mixing of condensable species, sedimentation, nucleation and coagulation. We find that the condensability of species is highly dependent on the oxygen abundance of an atmosphere. Oxygen poor atmospheres can be heated by UV and optical absorbers PS, TiO and CN which create temperature inversions. These inhibit condensation. Oxygen rich atmospheres are colder without temperature inversions, and are therefore more favourable environments for cloud formation. The major expected cloud component in O-dominated atmospheres with solar refractory abundance is TiO$_2$(s). Spectral features of clouds in these worlds are stronger in transmission than in emission, in particular at short wavelengths. We find a lack of optical data of solid species in comparison to the variety of stable cloud components which can form on hot, rocky planets.

astro-ph.EP↗

arXiv2025

GraphCoT-VLA: A 3D Spatial-Aware Reasoning Vision-Language-Action Model for Robotic Manipulation with Ambiguous Instructions

Vision-language-action models have emerged as a crucial paradigm in robotic manipulation. However, existing VLA models exhibit notable limitations in handling ambiguous language instructions and unknown environmental states. Furthermore, their perception is largely constrained to static two-dimensional observations, lacking the capability to model three-dimensional interactions between the robot and its environment. To address these challenges, this paper proposes GraphCoT-VLA, an efficient end-to-end model. To enhance the model's ability to interpret ambiguous instructions and improve task planning, we design a structured Chain-of-Thought reasoning module that integrates high-level task understanding and planning, failed task feedback, and low-level imaginative reasoning about future object positions and robot actions. Additionally, we construct a real-time updatable 3D Pose-Object graph, which captures the spatial configuration of robot joints and the topological relationships between objects in 3D space, enabling the model to better understand and manipulate their interactions. We further integrates a dropout hybrid reasoning strategy to achieve efficient control outputs. Experimental results across multiple real-world robotic tasks demonstrate that GraphCoT-VLA significantly outperforms existing methods in terms of task success rate and response speed, exhibiting strong generalization and robustness in open environments and under uncertain instructions.

cs.RO↗

arXiv2025

On Linking Planet Formation Models, Protoplanetary Disk Properties, and Mature Gas Giant Exoplanet Atmospheres

Measuring a single elemental ratio (e.g., carbon-to-oxygen) provides insufficient information for understanding the formation mechanisms and evolution that affect our observations of gas giant planet atmospheres. Although the fields of planet formation, protoplanetary disks, and exoplanets are well established and interconnected, our understanding of how to self-consistently and accurately link the theoretical and observational aspects of these fields together is lacking. To foster interdisciplinary conversations, the Max-Planck Institut f\"ur Astronomie (MPIA) hosted a week-long workshop called, "Challenge Accepted: Linking Planet Formation with Present-Day Atmospheres." Here, we summarize the latest theories and results in planet formation modeling, protoplanetary disk observations, and atmospheric observations of gas giant atmospheres to address one of the challenges of hosting interdisciplinary conferences: ensuring everyone is aware of the state-of-the-art results and technical language from each discipline represented. Additionally, we highlight key discussions held at the workshop. Our main conclusion is that it is unclear what the ideal observable is to make this link between formation scenarios and exoplanet atmospheres, whether it be multiple elemental abundance ratios, measuring refractory budgets, or something else. Based on discussions held throughout the workshop, we provide several key takeaways of what the workshop attendees feel need the most improvement and exploration within each discipline.

astro-ph.EP↗

arXiv2025

SpatialCoT: Advancing Spatial Reasoning through Coordinate Alignment and Chain-of-Thought for Embodied Task Planning

Spatial reasoning is an essential problem in embodied AI research. Efforts to enhance spatial reasoning abilities through supplementary spatial data and fine-tuning have proven limited and ineffective when addressing complex embodied tasks, largely due to their dependence on language-based outputs. While some approaches have introduced a point-based action space to mitigate this issue, they fall short in managing more intricate tasks within complex environments. This deficiency arises from their failure to fully exploit the inherent thinking and reasoning capabilities that are fundamental strengths of Vision-Language Models (VLMs). To address these limitations, we propose a novel approach named SpatialCoT, specifically designed to bolster the spatial reasoning capabilities of VLMs. Our approach comprises two stages: spatial coordinate bi-directional alignment, which aligns vision-language inputs with spatial coordinates, and chain-of-thought spatial grounding, which harnesses the reasoning capabilities of language models for advanced spatial reasoning. We evaluate SpatialCoT on challenging navigation and manipulation tasks, both in simulation and real-world settings. Experimental results demonstrate that our method significantly outperforms previous state-of-the-art approaches in both tasks.

cs.RO↗

arXiv2024

ExoLyn: a golden mean approach to multi-species cloud modelling in atmospheric retrieval

Context. Clouds are ubiquitous in exoplanets' atmospheres and play an important role in setting the opacity and chemical inventory of the atmosphere. Understanding clouds is a critical step in interpreting exoplanets' spectroscopic data. Aims. The aim is to model the multi-species nature of clouds in atmospheric retrieval studies. To this end, we develop ExoLyn - a 1D cloud model that balances physical consistency with computational efficiency. Methods. ExoLyn solves the transport equation of cloud particles and vapor under cloud condensation rates that are self-consistently calculated from thermodynamics. ExoLyn is a standalone, open source package capable to be combined with \texttt{optool} to calculate solid opacities and with \texttt{petitRADTRANS} to generate transmission or emission spectra. Results. With ExoLyn we find that the compositional structure of clouds in hot Jupiter planets' atmospheres is layered with a cloud dominated by magnesium-silicates on top of an iron cloud. This finding is consistent with more complex cloud formation models but can be obtained with ExoLyn in only a few seconds. The composition of the cloud particles can be constrained from the spectrum, for example, MgSiO3 and Mg2SiO4 components give rise to an absorption feature at 8 - 10 um. We investigate the dependence of the cloud structure on the bulk elemental composition of the planet and find that SiO2-dominated clouds forms on metal-rich planet and Fe clouds with strong extinction effect forms on C-rich planet. Conclusions. Designed towards maximum flexibility, ExoLyn can also be used in retrieval analysis of sub-Neptunes and self-luminous planets. The efficiency of ExoLyn opens the possibility of joint retrieval of exoplanets' gas and cloud components.

astro-ph.EP↗

arXiv2023

Accretion of aerodynamically large pebbles

Due to their aerodynamical coupling with gas, pebbles in protoplanetary discs can drift over large distances to support planet growth in the inner disc. In the past decade, this pebble accretion has been studied extensively for aerodynamically small pebbles (Stokes number St < 1). However, accretion can also operate in the St > 1 mode, e.g., when planetesimals collisionally fragment to smaller bodies or when the primordial gas disc disperses. This work aims to extend the study of pebble accretion to these aerodynamically loosely coupled particles. We integrate the pebble's equation of motion, accounting for gas drag, stellar and planetary gravity, in the midplane of a laminar disc. The accretion probability ($ε$) is calculated as function of Stokes number, disc pressure gradient index, planet mass and eccentricity. We find that for Stokes number above unity $ε$(St) first rises, due to lower drift and aided by a large atmospheric capture radius, until it reaches a plateau where the efficiency approaches 100 per cent. At high St the plateau region terminates as particles become trapped in resonance. These results are well described by a semi-analytical "kick-and-drift" model and we also provide fully analytical prescriptions for $ε$. We apply our model to the accretion of $\sim 30 μ$m dust particles in a dispersing protoplanetary and secondary (CO-rich) debris disc. It shows that physically small particles are mainly accreted as aerodynamically large Stokes number pebbles during the debris disc phase. Earth-mass planets may obtain $\sim 25$ per cent of their heavy elements through this late accretion phase.

astro-ph.EP↗