Searcharxiv⌕ Search

arXiv subjects

Taeyun Kim

Publications and source records attributed to Taeyun Kim.

10 recordsLinked to original sources

Which Taxels, Not How Many: Tactile Sensor Placement for Dexterous Policies

Policies for dexterous manipulation depend on design choices made before training: which sensors enter the observation, where they are placed, and how their readings are arranged for the encoder. These choices are rarely examined. For tactile sensing, simulation studies place taxels uniformly over the hand, and this placement has become a default without being tested. We test this default under a single controlled protocol. The protocol holds the teacher, the training procedure, and the candidate taxels fixed and varies only the sensor configuration and the representation. We evaluate it on in-palm rotation and grasping with three robot hands. A small set of taxels selected from contact statistics matches a far larger uniform set. Taxels at rarely contacted sites are equivalent to no tactile input. With all candidate taxels, their layout in the input tensor does not affect performance, even for a convolutional encoder. Uniform placement therefore overestimates the number of taxels a task requires. We distill these results into lessons for designing tactile policies and into a minimal set of design choices that a tactile policy study should report.

cs.RO↗

SkateNav: Learning Energy-Efficient Navigation through Adaptive Skating on Passive-Wheel Quadrupedal Robots

Passive-wheel quadrupedal robots offer energy-efficient locomotion through leg-driven propulsion and momentum-driven gliding. However, navigation requires coordinating steering with intermittent propulsion and adapting push-glide timing to task demands. Velocity-tracking objectives may suppress the speed variations inherent in skating, discouraging efficient gliding. To address these challenges, we propose SkateNav, a unified skating-navigation framework based on reinforcement learning. A time-budgeted goal-reaching objective allows the policy to jointly learn steering, propulsion, and their temporal allocation. An adaptive skating cycle structures gliding, pushing, and recovery, while frequency modulation at every control step adjusts phase durations to navigation demands. Simulation experiments demonstrate higher average success rates and lower cost of transport than the evaluated planner-controller baselines, with ablations confirming the contributions of cycle structure and frequency adaptation. Real-world experiments demonstrate zero-shot transfer from simulation to a passive-wheel Mini Cheetah across cluttered indoor spaces, slopes, and paved outdoor surfaces. Supplementary videos are available on the project page: https://ktw1404.github.io/SkateNav/

cs.RO↗

NeuDonatello: Uncertainty-Aware Framework for Accurate Neural SDF Learning

Neural surface reconstruction has emerged as a powerful paradigm for recovering high-quality 3D surfaces from multi-view images. However, recovering accurate geometry solely from RGB images remains challenging due to uncertainties arising from textureless regions, occlusions, and inherent scene ambiguities. Existing methods often overlook such uncertainties, leading to inaccurate estimates of the signed distance function (SDF). We introduce NeuDonatello, a novel framework that models and leverages SDF uncertainty to improve surface reconstruction. Central to our approach is to model spatially varying uncertainty using a Monte Carlo sampling strategy. Using this uncertainty, we develop an adaptive regularization that selectively strengthens geometric constraints where RGB supervision is unreliable, avoiding incorrect surface reconstruction. We further introduce an uncertainty-aware scale parameter for the SDF-to-density conversion. Conditioned on uncertainty, this design enables more accurate modeling of spatially varying densities. Extensive experiments demonstrate that NeuDonatello achieves state-of-the-art reconstruction accuracy, with robust performance across diverse scenes using only posed RGB images.

cs.CV↗

Multistage development of short-range spin correlations and weak magnetic order in the two coupled trillium lattices of K2Fe2(MoO4)(PO4)2

Trillium lattices, where magnetic ions form a chiral network of corner-sharing triangles, offer a three-dimensional magnetic framework that can host fragile classical spin-liquid states. Herein, we report on the magnetization, specific heat, electron spin resonance (ESR), and muon spin relaxation ($μ$SR) of K$_{2}$Fe$_{2}$(MoO$_{4}$)(PO$_{4}$)$_{2}$ single crystals. Magnetization measurements reveal strong antiferromagnetic interactions coexisting with weak magnetic order at $T_{\rm N} = 5.2$~K, as evidenced by a $λ$-like anomaly observed in the magnetic susceptibility, a critical enhancement of the muon spin relaxation rate and the wipeout of the ESR signal as the temperature approaches $T_{\rm N}$. Above $T_{\rm N}$, two distinct developments of short-range spin correlations are identified at $T_{\rm H} = 34$~K and $T_{\rm L} = 10$~K, supported by magnetic specific heat anomalies and the temperature dependence of the ESR linewidth and $g$-factor. Upon cooling below $T_{\rm N}$, an anomaly appears at $T^{*} = 3.2$~K in thermodynamic observables and the muon spin relaxation rate, indicative of spin reorientation driven by residual interactions. Despite the presence of magnetic order, $μ$SR experiments reveal dynamically fluctuating spins persisting even in the ordered state. Moreover, the suppression of $T_{\rm N}$ under applied magnetic fields ($μ_{0}H \geq 2$~T) suggests that K$_{2}$Fe$_{2}$(MoO$_{4}$)(PO$_{4}$)$_{2}$ constitutes a promising candidate for exploring field-induced spin-liquid behavior in three-dimensionally coupled trillium lattices.

cond-mat.str-el↗

CLUE: Adaptively Prioritized Contextual Cues by Leveraging a Unified Semantic Map for Effective Zero-Shot Object-Goal Navigation

Zero-shot object-goal navigation (ZSON) is a challenging problem in robotics that requires a comprehensive understanding of both language and visual observations. Contextual cues from rooms and objects are critical, but their relative importance depends on the target: some objects are strongly tied to specific room types, while others are better predicted by nearby co-located objects. Existing methods overlook this distinction, leading to inefficient and inaccurate exploration. We present CLUE, a novel navigation framework that adaptively balances the use of contextual rooms and objects by leveraging commonsense knowledge extracted from an offline large language model (LLM). By estimating a target's association with room types using LLM, the agent prioritizes room cues for predictable objects and object cues for those with weak room associations. Our framework constructs a unified semantic value map that integrates both types of contextual information, adaptively weighted by the target's ambiguity to guide exploration. Combined with multi-viewpoint verification and an exploration strategy informed by contextual cues, CLUE achieves robust and efficient navigation. Extensive experiments in simulation and real-world deployments show that our method consistently outperforms state-of-the-art baselines in both success rate (SR) and success weighted by path length (SPL), demonstrating its effectiveness and practicality for real-world navigation tasks.

cs.RO↗

$5/9-$Magnetization Plateau and Spin Supersolidity in YCu$_3$(OD)$_{7-x}$Br$_{2+x}$ under Magnetic Fields up to 120~T

We performed high-precision magnetization measurements up to 120~T on three compositions of the newly discovered kagome antiferromagnet YCu$_3$(OD)$_{7-x}$Br$_{2+x}$ (YCOB), revealing a previously unobserved 5/9 fractional magnetization plateau. All YCOB samples with different Br$^-$ concentrations exhibit nearly identical magnetization curves below 60~T, whereas the 5/9 plateau appears at markedly different fields in the ultrahigh-field regime. By modeling the experimental data using tensor-network calculations, we derive the effective spin Hamiltonians for the YCOB family with three spatially anisotropic Heisenberg couplings (the 3$J$-type model), which quantitatively reproduces the measured magnetization processes and captures the composition-dependent evolution of the 5/9 plateau. Furthermore, our theoretical analysis suggests the emergence of a spin supersolid phase in the field window between the 1/3 and 5/9 plateaus, which is sensitive to spin exchange parameters and accounts for the significant variation in the critical fields of the 5/9 plateau observed among different YCOB compositions.

cond-mat.str-el↗

Multistage spin correlations in the $s$ = 1/2 stuffed hyper-star lattice Li$_{2}$Cu$_{2}$(MoO$_{4}$)$_{3}$

Star lattice, which can be visualized as a honeycomb network with each vertex replaced by a triangle, provides a rare platform for realizing exotic quantum states such as quantum spin liquids and disorder-driven random-singlet (RS) states. Herein, we investigate the ground-state properties of the three-dimensional (3D) stuffed hyper-star lattice Li$_2$Cu$_2$(MoO$_4$)$_3$, which exhibits a crossover from short-range spin correlations to a disorder-driven RS-like state below $T^{*}\sim$15.8 K. Thermodynamic and microscopic measurements capture this crossover through a change in the power-law behavior of various observables, from $\sim T^{0.25}$ for $T > T^{*}$ to $\sim T^{-0.50}$ for $T < T^{*}$. Upon further cooling, a quasi-frozen state emerges near $T_{\rm f} = 0.32$ K, likely associated with weakly coupled spin chains within the hyper-star spin network. Our results underscore the crucial role of orphan spins and weak residual interactions in stabilizing a disorder-driven quantum-disordered state in 3D.

cond-mat.str-el↗

Structural Incompatibility of Differentiable Sorting and Within-Vector Rank Normalization

We show that differentiable sorting and ranking operators are structurally incompatible with within-vector rank normalization. We formalize admissibility through monotone invariance (C1), batch independence (C2), and a rank-space stability condition (C3). Gap-sensitive relaxations such as SoftSort violate (C1) by a quantitative margin that depends on the temperature and input scale. Batchwise rank relaxations such as SinkhornSort violate (C2): the same sample can be assigned outputs arbitrarily close to 0 or 1 depending solely on batch context. Condition (C3) implies (C1) under the rank representation used here and should not be read as a third independent failure mode. We also characterize the admissible class: any admissible operator must factor through the rank representation via a Lipschitz function.

cs.LG↗

Coexistence of anomalous spin dynamics and weak magnetic order in a chiral trillium lattice K2FeSn(PO4)3

Trillium lattices, where magnetic ions form a three-dimensional chiral network of corner-sharing equilateral triangular motifs, offer a prominent platform to explore exotic quantum states. In this work, we report ground-state properties of the $S$ = 5/2 trillium lattice compound K$_{2}$FeSn(PO$_{4}$)$_{3}$ through thermodynamic, electron spin resonance (ESR), and muon spin relaxation ($μ$SR) experiments. Thermodynamic and ESR measurements reveal the two-step evolution of magnetic correlations across $T^{*}$ = 11 K, which results from an interplay between dominant antiferromagnetic Heisenberg interactions and subleading interactions. Below $T^{*}$, \textit{dc} and \textit{ac} magnetic susceptibilities indicate weak \textcolor{black}{magnetic ordering} at $T_{\rm N} \approx 2$ K under low fields, which is suppressed for $μ_{0}H \geq 2$ T, consistent with a power-law dependence of magnetic specific heat at low temperatures. $μ$SR experiments confirm the dominance of persistent spin dynamics and the absence of conventional spin freezing, supporting the subtle nature of weak magnetic ordering coexisting with spin-liquid-like fluctuations. These findings underscore the potential for realizing a classical spin-liquid ground state with exotic excitations in high-spin trillium lattice systems.

cond-mat.str-el↗

SaWa-ML: Structure-Aware Pose Correction and Weight Adaptation-Based Robust Multi-Robot Localization

Multi-robot localization is a crucial task for implementing multi-robot systems. Numerous researchers have proposed optimization-based multi-robot localization methods that use camera, IMU, and UWB sensors. Nevertheless, characteristics of individual robot odometry estimates and distance measurements between robots used in the optimization are not sufficiently considered. In addition, previous researches were heavily influenced by the odometry accuracy that is estimated from individual robots. Consequently, long-term drift error caused by error accumulation is potentially inevitable. In this paper, we propose a novel visual-inertial-range-based multi-robot localization method, named SaWa-ML, which enables geometric structure-aware pose correction and weight adaptation-based robust multi-robot localization. Our contributions are twofold: (i) we leverage UWB sensor data, whose range error does not accumulate over time, to first estimate the relative positions between robots and then correct the positions of each robot, thus reducing long-term drift errors, (ii) we design adaptive weights for robot pose correction by considering the characteristics of the sensor data and visual-inertial odometry estimates. The proposed method has been validated in real-world experiments, showing a substantial performance increase compared with state-of-the-art algorithms.

cs.RO↗