arXiv subjects

Xiaohai Hu

Publications and source records attributed to Xiaohai Hu.

7 recordsLinked to original sources

arXiv2026

Decomposing Predictive Kubernetes Autoscaling for Large Language Model Serving Under Long Startup Delays

Large language model (LLM) inference deployed on Kubernetes faces an autoscaling challenge that conventional web services do not: new serving replicas take two to ten minutes to start because multi-gigabyte model weights must be loaded, which makes purely reactive scaling structurally late. We ask a sharp question: among the components of a predictive autoscaler, which ones actually matter under such long actuation delays? We answer it by decomposing predictive autoscaling into four factors---token-aware demand tracking, startup-delay lookahead, a bounded uncertainty margin, and plant-state observation---and measuring each factor's contribution in isolation on production-derived heavy-tailed workloads generated by ServeGen. Our main finding is that a simple exponentially weighted moving average (EWMA) predictor with delay-aware lookahead and an upper confidence bound (UCB) margin captures most of the benefit, reducing time-to-first-token (TTFT) service-level-objective (SLO) violations from 53% (reactive, queries-per-second based) to 0.5% across five random seeds; lookahead alone is the single largest factor, a 14$\times$ reduction. Kalman filter variants do not consistently improve the cost--SLO tradeoff. Controlled experiments isolate the reason token granularity is necessary: context length, through key--value cache pressure, degrades TTFT far more than request rate at matched throughput. Finally, a validation on a real Kubernetes cluster (Qwen2.5-7B, A100, vLLM) confirms the central mechanism: a delay-aware lookahead controller cuts TTFT violations from 63.5\% to 3.7\% relative to reactive KEDA scaling. We distinguish throughout which findings are specific to Kubernetes actuation and which are general to LLM serving.

cs.PF↗

arXiv2026

MARVL: Multi-Stage Guidance for Robotic Manipulation via Vision-Language Models

Designing dense reward functions is pivotal for efficient robotic Reinforcement Learning (RL). However, most dense rewards rely on manual engineering, which fundamentally limits the scalability and automation of reinforcement learning. While Vision-Language Models (VLMs) offer a promising path to reward design, naive VLM rewards often misalign with task progress, struggle with spatial grounding, and show limited understanding of task semantics. To address these issues, we propose MARVL-Multi-stAge guidance for Robotic manipulation via Vision-Language models. MARVL fine-tunes a VLM for spatial and semantic consistency and decomposes tasks into multi-stage subtasks with task direction projection for trajectory sensitivity. Empirically, MARVL significantly outperforms existing VLM-reward methods on the Meta-World benchmark, demonstrating superior sample efficiency and robustness on sparse-reward manipulation tasks.

cs.RO↗

arXiv2025

Iterative Youla-Kucera Loop Shaping For Precision Motion Control

This paper presents a numerically robust approach to multi-band disturbance rejection using an iterative Youla-Kucera parameterization technique. The proposed method offers precise control over shaping the frequency response of a feedback loop while maintaining numerical stability through a systematic design process. By implementing an iterative approach, we overcome a critical numerical issue in rejecting vibrations with multiple frequency bands. Meanwhile, our proposed modification of the all-stabilizing Youla-Kucera architecture enables intuitive design while respecting fundamental performance trade-offs and minimizing undesired waterbed amplifications. Numerical validation on a hard disk drive servo system demonstrates significant performance improvements, enabling enhanced positioning precision for increased storage density. The design methodology extends beyond storage systems to various high-precision control applications where multi-band disturbance rejection is critical.

eess.SY↗

arXiv2024

Learned Slip-Detection-Severity Framework using Tactile Deformation Field Feedback for Robotic Manipulation

Safely handling objects and avoiding slippage are fundamental challenges in robotic manipulation, yet traditional techniques often oversimplify the issue by treating slippage as a binary occurrence. Our research presents a framework that both identifies slip incidents and measures their severity. We introduce a set of features based on detailed vector field analysis of tactile deformation data captured by the GelSight Mini sensor. Two distinct machine learning models use these features: one focuses on slip detection, and the other evaluates the slip's severity, which is the slipping velocity of the object against the sensor surface. Our slip detection model achieves an average accuracy of 92%, and the slip severity estimation model exhibits a mean absolute error (MAE) of 0.6 cm/s for unseen objects. To demonstrate the synergistic approach of this framework, we employ both the models in a tactile feedback-guided vertical sliding task. Leveraging the high accuracy of slip detection, we utilize it as the foundational and corrective model and integrate the slip severity estimation into the feedback control loop to address slips without overcompensating.

cs.RO↗

arXiv2024

Context and Geometry Aware Voxel Transformer for Semantic Scene Completion

Vision-based Semantic Scene Completion (SSC) has gained much attention due to its widespread applications in various 3D perception tasks. Existing sparse-to-dense approaches typically employ shared context-independent queries across various input images, which fails to capture distinctions among them as the focal regions of different inputs vary and may result in undirected feature aggregation of cross-attention. Additionally, the absence of depth information may lead to points projected onto the image plane sharing the same 2D position or similar sampling points in the feature map, resulting in depth ambiguity. In this paper, we present a novel context and geometry aware voxel transformer. It utilizes a context aware query generator to initialize context-dependent queries tailored to individual input images, effectively capturing their unique characteristics and aggregating information within the region of interest. Furthermore, it extend deformable cross-attention from 2D to 3D pixel space, enabling the differentiation of points with similar image coordinates based on their depth coordinates. Building upon this module, we introduce a neural network named CGFormer to achieve semantic scene completion. Simultaneously, CGFormer leverages multiple 3D representations (i.e., voxel and TPV) to boost the semantic and geometric representation abilities of the transformed 3D volume from both local and global perspectives. Experimental results demonstrate that CGFormer achieves state-of-the-art performance on the SemanticKITTI and SSCBench-KITTI-360 benchmarks, attaining a mIoU of 16.87 and 20.05, as well as an IoU of 45.99 and 48.07, respectively. Remarkably, CGFormer even outperforms approaches employing temporal images as inputs or much larger image backbone networks.

cs.CV↗

arXiv2024

Learning to Detect Slip through Tactile Estimation of the Contact Force Field and its Entropy

Detection of slip during object grasping and manipulation plays a vital role in object handling. Existing solutions primarily rely on visual information to devise a strategy for grasping. However, for robotic systems to attain a level of proficiency comparable to humans, especially in consistently handling and manipulating unfamiliar objects, integrating artificial tactile sensing is increasingly essential. We introduce a novel physics-informed, data-driven approach to detect slip continuously in real time. We employ the GelSight Mini, an optical tactile sensor, attached to custom-designed grippers to gather tactile data. Our work leverages the inhomogeneity of tactile sensor readings during slip events to develop distinctive features and formulates slip detection as a classification problem. To evaluate our approach, we test multiple data-driven models on 10 common objects under different loading conditions, textures, and materials. Our results show that the best classification algorithm achieves a high average accuracy of 95.61%. We further illustrate the practical application of our research in dynamic robotic manipulation tasks, where our real-time slip detection and prevention algorithm is implemented.

cs.RO↗

arXiv2023

GraphLLM: Boosting Graph Reasoning Ability of Large Language Model

The advancement of Large Language Models (LLMs) has remarkably pushed the boundaries towards artificial general intelligence (AGI), with their exceptional ability on understanding diverse types of information, including but not limited to images and audio. Despite this progress, a critical gap remains in empowering LLMs to proficiently understand and reason on graph data. Recent studies underscore LLMs' underwhelming performance on fundamental graph reasoning tasks. In this paper, we endeavor to unearth the obstacles that impede LLMs in graph reasoning, pinpointing the common practice of converting graphs into natural language descriptions (Graph2Text) as a fundamental bottleneck. To overcome this impediment, we introduce GraphLLM, a pioneering end-to-end approach that synergistically integrates graph learning models with LLMs. This synergy equips LLMs with the ability to proficiently interpret and reason on graph data, harnessing the superior expressive power of graph learning models. Our empirical evaluations across four fundamental graph reasoning tasks validate the effectiveness of GraphLLM. The results exhibit a substantial average accuracy enhancement of 54.44%, alongside a noteworthy context reduction of 96.45% across various graph reasoning tasks.

cs.CL↗