arXiv subjects

Haiyan Mao

Publications and source records attributed to Haiyan Mao.

4 recordsLinked to original sources

arXiv2026

UMI-Bridge: Action-Anchored Latent Alignment across Human and Robot Manipulation Data

Real-robot demonstrations are limited, motivating the use of human manipulation data collected without robots, including egocentric videos and handheld Universal Manipulation Interface (UMI) demonstrations. However, differences in viewpoint, embodiment, and available action supervision make it difficult to align representations across these sources according to manipulation motion rather than visual appearance. We introduce UMI-Bridge, which uses UMI as an intermediate domain to align representations according to action equivalence rather than pixel similarity. UMI action supervision anchors the latent representation to end-effector motion and gripper behavior, while synchronized head-wrist observations and paired ego-UMI clips support alignment across views and domains. We train a dual-view latent action model (LAM) on human manipulation data without robot demonstrations, then freeze its wrist teacher and dynamics model to regularize vision-language-action (VLA) post-training on UMI and robot data. The shared wrist interface enables this training-time supervision across both domains while preserving the policy's standard inference architecture. Across three real-robot tasks, UMI-Bridge achieves 91.7% mean success versus 73.3% for Naive Co-training with matched UMI and robot data. On two data-efficiency tasks, it surpasses a full-data Robot-only baseline using 25% of the robot demonstrations together with UMI data. It also achieves 85% and 90% success on two additional tasks learned from UMI demonstrations without task-specific robot demonstrations. These results support action-anchored latent alignment for data-efficient robot learning and UMI-to-robot task transfer.

cs.RO↗

arXiv2026

SEA-Nav: Efficient Policy Learning for Safe and Agile Quadruped Navigation in Cluttered Environments

Efficiently learning safe and agile quadruped navigation in densely cluttered environments remains difficult: existing methods often lack safety and agility, or become conservative in complex scenes and require long training schedules. We propose SEA-Nav (Safe, Efficient, and Agile Navigation), a safe reinforcement learning framework for quadruped navigation in cluttered environments. A differentiable control barrier function (CBF) shield constrains the policy to produce safe velocity commands. An adaptive collision-state initialization mechanism increases the probability of learning from safety-critical near-collision experience. An action regularization term further suppresses infeasible commands for physical deployment. The policy converges after about one hour of training on a single RTX 4090 and transfers zero-shot to real-world cluttered scenes.

cs.RO↗

arXiv2025

Out-of-time-order correlators bridge classical transport and quantum dynamics

The out-of-time-order correlator (OTOC) has emerged as a central tool for quantifying decoherence across wide-ranging physical platforms. Here we demonstrate its direct measurement in a classical ensemble using nuclear magnetic resonance (NMR) with a modulated gradient spin echo (MGSE) sequence and extend the method into a multidimensional correlation to track exchange phenomena. Position is encoded through magnetic field gradients and momentum through the velocity autocorrelation function, enabling experimental access to OTOCs for proton motion confined within the self-similar lattice of the metal-organic framework MOF-808. Here, water confined to specified geometries within the MOF pores gives rise to spatially distinct diffusive eigenmodes with characteristic relative entropies. We demonstrate that periodic radiofrequency (rf) driving combined with gradient modulation yields entropy evolution through the selection of distinct diffusion modes. Frequency-resolved diffusion spectra connect these entropy dynamics to classical heat-exchange laws, revealing how operational features of quantum systems are mirrored in confined, macroscopic spin ensembles.

quant-ph↗

arXiv2019

Double Perovskite Structure Induced by Co Addition to PbTiO$_3$ : Insights from DFT and Experimental Solid State NMR Spectroscopy

The effects of Co addition on the chemical and electronic structure of PbTiO$_3$ were explored both by theory and through experiment. Cobalt was incorporated to PbTiO$_3$ during sol gel process. The XRD data of the compounds confirmed the perovskite structure for the pure samples. The XRD lines broadened and showed emerging cubic-like features as the Co incorporation increased. The changes in the XRD pattern were interpreted as double perovskite structure formation. $^{207}$Pb NMR measurements revealed a growing isotropic component in the presence of Co. In line with the experiments, DFT calculated chemical-shift values corroborate isotropic coordination of Pb suggesting the formation of cubic Pb$_2$CoTiO$_6$ domains in the prepared samples. The state-of-the-art hybrid functional first-principles calculations indicate formation of Pb$_2$CoTiO$_6$ with cubic structure and confirms that Co addition can decrease oxygen binding energy significantly. Experimental UV-Vis spectroscopy results indicate that upon addition of Co, the band gap is shifted towards visible wavelengths which was confirmed by the energy bands and absorption spectra calculations. The oxygen binding energies were determined by temperature programmed reduction (TPR) measurements. Upon addition of Co, TPR lines shifted to lower temperatures and new features appeared in the TPR patterns. This shift was interpreted as weakening of oxygen cobalt bond strength. The change in the electronic structure by the alterations of oxygen vacancy formation energy and bond lengths upon Co insertion are determined by DFT calculations.

cond-mat.str-el↗