arXiv subjects

Zhang Xiao

Publications and source records attributed to Zhang Xiao.

4 recordsLinked to original sources

arXiv2026

Hybrid Orchestration of Edge AI and Microservices via Graph-based Self-Imitation Learning

Modern edge AI applications increasingly rely on microservice architectures that integrate both AI services and conventional microservices into complex request chains with stringent latency requirements. Effectively orchestrating these heterogeneous services is crucial for ensuring low-latency performance, yet remains challenging due to their diverse resource demands and strong operational interdependencies under resource-constrained edge environments. In particular, frequent interactions between services tightly couple deployment and routing decisions, yet existing approaches optimize them in isolation, leading to fundamentally inadequate system performance.In this paper, we propose SIL-GPO, a reinforcement learning framework that optimizes hybrid orchestration for edge AI microservice systems. SIL-GPO formulates the orchestration problem as a sequential decision-making task and leverages graph attention networks to encode service topologies and routing dependencies within the agent state representation. Moreover, SIL-GPO integrates a self-imitation learning strategy into proximal policy optimization, enabling the agent to prioritize and reuse high-reward trajectories. This guides policy updates towards globally promising solutions that standard RL often fails to discover under sparse rewards and large combinatorial action spaces. We conduct extensive experiments on trace-driven edge AI workloads, demonstrating that SIL-GPO significantly reduces end-to-end service latency and enhances resource utilization compared to state-of-the-art heuristic, metaheuristic, and deep RL baselines. Our framework offers a unified and scalable solution for efficient orchestration of AI services and microservices in the edge, paving the way for low-latency, high-performance edge AI deployments.

cs.NI↗

arXiv2026

Optimization of Higher-Order Harmonic Surface Tessellations for Additively Manufactured Air-to-Air Heat Exchangers

Air-to-air heat exchangers are vital for energy recovery and thermal management but often suffer from reduced effectiveness, high pressure losses, and increased pumping power in conventional designs. Advances in additive manufacturing have enabled nature-inspired geometries, such as lattice and triply periodic minimal surface (TPMS) structures, which enhance heat transfer through complex first-order surfaces but frequently cause excessive pressure drops. This study proposes an optimized higher-order harmonic heat-transfer surface tessellation developed through an optimization framework integrating analytical and numerical methods. The goal is to improve the overall thermal-hydraulic performance of the heat exchanger over a range of operating conditions. Results of sensitivity analysis show that secondary surface modification of this type can yield significant increase in the effectiveness reaching up to 70% although with associated increase in the pressure drop. The secondary surface wave frequency was found to be a more important control parameter than the amplitude in achieving high thermal-hydraulic performance. Additionally, we show that the optimized second order harmonic-type structure achieved relatively higher effectiveness and lower pressure-drop than the gyroid structure in the turbulent flow regime for Re>=7000. Although the gyroid TPMS structure had relatively higher effectiveness in the laminar and weakly turbulent flow regime, the associated pressure drop was found to be significantly higher than that of the harmonic-type structure.

physics.flu-dyn↗

arXiv2024

A real-time, robust and versatile visual-SLAM framework based on deep learning networks

This paper explores how deep learning techniques can improve visual-based SLAM performance in challenging environments. By combining deep feature extraction and deep matching methods, we introduce a versatile hybrid visual SLAM system designed to enhance adaptability in challenging scenarios, such as low-light conditions, dynamic lighting, weak-texture areas, and severe jitter. Our system supports multiple modes, including monocular, stereo, monocular-inertial, and stereo-inertial configurations. We also perform analysis how to combine visual SLAM with deep learning methods to enlighten other researches. Through extensive experiments on both public datasets and self-sampled data, we demonstrate the superiority of the SL-SLAM system over traditional approaches. The experimental results show that SL-SLAM outperforms state-of-the-art SLAM algorithms in terms of localization accuracy and tracking robustness. For the benefit of community, we make public the source code at https://github.com/zzzzxxxx111/SLslam.

cs.RO↗

arXiv2020

Bound states in the continuum are universal under the effect of minimal length

Bound states in the continuum (BICs) are generally considered unusual phenomena. In this work, we provide a method to analyze the spatial structure of particle's bound states in the presence of a minimal length, which can be used to find BICs. It is shown that the BICs are universal phenomena under the effect of the minimal length. Several examples of typical potentials, i.e., infinite potential well, linear potential, harmonic oscillator, quantum bouncer and Coulomb potential, et al, are provided to show the BICs are universal. The wave functions and energy of the first three examples are provided. A condition is obtained to determine whether the BICs can be readily found in systems. Using the condition, we find that although the BICs are universal phenomena, they are often hardly found in many ordinary environments since the bound continuous states perturbed by the effect of the minimal length are too weak to observe. The results are consistent with the current experimental results on BICs. In addition, we reveal a mechanism of the BICs. The mechanism explains why current research shows the bound discrete states are typical, whereas BICs are always found in certain particular environments when the minimal length is not considered.

quant-ph↗