arXiv subjects

Li Peng

Publications and source records attributed to Li Peng.

36 records · Page 2Linked to original sources

arXiv2025

Voltage Synchronization and Proportional Current Sharing of Grid-Forming Inverters

Most previously proposed controllers are analyzed in the small-signal/quasi-steady regime rather than large-signal or transient stability for grid-forming inverters (GFMI). Additionally, methods that presume system-wide data--global measurements and complete grid-model knowledge--are challenging to realize in practice and unsuitable for large-scale operation. Moreover, proportional current sharing is rarely embedded into them. The whole system is a high-order, nonlinear differential system, making analysis intractable without principled simplifications. Hence, contraction stability analysis in GFMI is proposed to guarantee the large-signal stability. Furthermore, a contraction-based controller is proposed to synchronize GFMI. Additionally, this paper proposes integrating an auxiliary virtual-impedance layer into the contraction-based controller to achieve proportional current sharing, while the GFMI retains global stability and voltage synchronization. A dispatchable virtual oscillator control (dVOC), also known as the Andronov--Hopf oscillator (AHO) is used to validate the proposed contraction stability analysis and contraction-based controller with virtual-impedance. It is proved that the complex multi-converter system can achieve output-feedback contraction under large-signal operation. Therefore, without requiring system-wide data, the proposed method offers voltage synchronization, decentralized stability conditions for the transient stability of AHO and proportional current sharing, beyond prior small-signal, quasi-steady analysis.

eess.SY↗

arXiv2025

Theoretical Grid-Forming Extreme of Inverters

What are the theoretical and physical limits of a grid-forming inverter? This letter proposes that the extreme grid-forming ability of inverters is limited by their dc-side, ac-side, circuit topology dynamics, but not control. While many papers focus on how to improve grid-forming inverters stability, power sharing, inertia emulation, fault response, few, if any, formally define the fundamental theoretical limits or extremes of grid-forming behavior. It seems that the grid-forming can be improved endlessly. No physical system can support a grid indefinitely without limitations, especially under increasing levels of disturbance or uncertainty. Therefore, this boundary is explicitly shown by a mathematical expression in this letter. Consequently, the results show that relatively low dc-side voltage and high active power injection could damage the grid-forming ability. Poor consideration of dc-side, ac-side, and circuit topology dynamics in real practice will cause jeopardizing oscillation even by the theoretical best grid-forming control strategy.

eess.SY↗

arXiv2025

Symmetric Sliding-Mode Control of Grid-Forming Inverters With Precision Region Under AC and DC Sides Varying

Voltage regulation under conventional grid-forming controllers is tightly coupled to power sharing and dc-link dynamics. Consequently, its tracking accuracy deteriorates during grid faults, sudden power sharing changes, or dc-bus voltage varying. To address this issue, a symmetric sliding-mode control (SSMC) method is developed and its voltage precision region is derived. It illustrates how much ac-side power dynamics and dc-link voltage varying can be decoupled from the voltage regulation task, which helps predict when an abnormal entangling appears. While conventional sliding-mode controls address voltage-tracking error through complex sliding surface designs, repetitive correction techniques or special reaching laws, this work identifies that the error at power-line frequency primarily stem from the asymmetry property of inverters with the delay effect and the computational inaccuracy. Guided by this insight, an asymmetry compensation structure is proposed, which avoids added design complexity and directly mitigates voltage tracking error. Furthermore, the control design is supported by a physical and quantitative explanation, aiding in parameter tuning. Simulation and experimental results demonstrate that the proposed method achieves faster tracking responses while maintaining robust and more accurate tracking under both dc-link voltage and ac-side current variations. Conventional grid-forming and classical sliding-mode controllers, which handle these variations separately, cannot match this combined speed and robustness. Furthermore, the voltage precision region is explicitly verified.

eess.SY↗

arXiv2025

Confidence-Guided Human-AI Collaboration: Reinforcement Learning with Distributional Proxy Value Propagation for Autonomous Driving

Autonomous driving promises significant advancements in mobility, road safety and traffic efficiency, yet reinforcement learning and imitation learning face safe-exploration and distribution-shift challenges. Although human-AI collaboration alleviates these issues, it often relies heavily on extensive human intervention, which increases costs and reduces efficiency. This paper develops a confidence-guided human-AI collaboration (C-HAC) strategy to overcome these limitations. First, C-HAC employs a distributional proxy value propagation method within the distributional soft actor-critic (DSAC) framework. By leveraging return distributions to represent human intentions C-HAC achieves rapid and stable learning of human-guided policies with minimal human interaction. Subsequently, a shared control mechanism is activated to integrate the learned human-guided policy with a self-learning policy that maximizes cumulative rewards. This enables the agent to explore independently and continuously enhance its performance beyond human guidance. Finally, a policy confidence evaluation algorithm capitalizes on DSAC's return distribution networks to facilitate dynamic switching between human-guided and self-learning policies via a confidence-based intervention function. This ensures the agent can pursue optimal policies while maintaining safety and performance guarantees. Extensive experiments across diverse driving scenarios reveal that C-HAC significantly outperforms conventional methods in terms of safety, efficiency, and overall performance, achieving state-of-the-art results. The effectiveness of the proposed method is further validated through real-world road tests in complex traffic conditions. The videos and code are available at: https://github.com/lzqw/C-HAC.

cs.RO↗

arXiv2025

Resource-Efficient Motion Control for Video Generation via Dynamic Mask Guidance

Recent advances in diffusion models bring new vitality to visual content creation. However, current text-to-video generation models still face significant challenges such as high training costs, substantial data requirements, and difficulties in maintaining consistency between given text and motion of the foreground object. To address these challenges, we propose mask-guided video generation, which can control video generation through mask motion sequences, while requiring limited training data. Our model enhances existing architectures by incorporating foreground masks for precise text-position matching and motion trajectory control. Through mask motion sequences, we guide the video generation process to maintain consistent foreground objects throughout the sequence. Additionally, through a first-frame sharing strategy and autoregressive extension approach, we achieve more stable and longer video generation. Extensive qualitative and quantitative experiments demonstrate that this approach excels in various video generation tasks, such as video editing and generating artistic videos, outperforming previous methods in terms of consistency and quality. Our generated results can be viewed in the supplementary materials.

cs.CV↗

arXiv2025

Mell: Memory-Efficient Large Language Model Serving via Multi-GPU KV Cache Management

Serving large language models (LLMs) for massive users is challenged by the significant memory footprint of the transient state, known as the key-value (KV) cache, which scales with sequence length and number of requests. Instead of renting or buying more expensive GPUs, the load imbalance of the KV cache across GPUs, coupled with recent advances in inter-GPU communication, provides an opportunity to serve more requests via request migration. However, high migration overhead and unpredictable request patterns make it challenging. Therefore, this paper proposes MELL, a memory-efficient LLM serving system via multi-GPU KV cache management. It saves the number of GPUs needed in the system by considering the dynamic KV cache load and the costly request migration. Specifically, we first develop an adaptive request migration mechanism to balance the computational and communication overheads and adapt to diverse resource conditions. Then, we design an online algorithm tailored to a multi-LLM request and multi-GPU scheduling problem with migration enabled. It aims to minimise the required GPUs while limiting the number of migrations. Finally, we implement a prototype of MELL and demonstrate that it reduces the number of GPUs by 31% and increases the GPU utilization by 43% at most compared to existing LLM serving systems.

cs.DC↗

arXiv2024

Unveiling quantum phases in quasi-one-dimensional dipolar gases using continuous matrix product state

We investigate the ground-state properties of the quasi-one-dimensional dipolar gases using continuous matrix product states techniques. Making use of the first- and second-order correlation functions, we find that the system supports the superfluid, super-Tonks-Girardeau, and quasicrystal phases according to the Luttinger liquid theory. We also map out the phase diagram on the parameter plane consisting the contact and dipolar interaction strengths. Furthermore, we compute the Luttinger parameter, the structure factor, and the momentum distribution of the system. Finally, we show that the predicted dipolar effect can potentially be observed in quasi-one-dimensional gases of polar molecules.

cond-mat.quant-gas↗

arXiv2024

Dimensional Analysis Theory and Molecular Dynamics Simulation of Polypropylene Melt Flow during Injection Molding Process

Flow marks are common surface defects that occur in injection-molded products. Their formation may be related to the flow process of the melt in the mold. Through dimensional analysis, we have discovered that the geometric shape of the flow field is controlled by specific dimensionless quantities. These quantities can be summarized as follows: geometric dimensionless quantities related to the shape of the mold, material dimensionless quantities related to the melt and mold materials, and physical dimensionless quantities related to the flow. When the geometric shape of the mold changes proportionally, with the melt and mold material fixed, and the initial temperature of the melt and mold fixed, the geometric shape of the flow field will be solely controlled by the Weissenberg number Wi. If Wi is kept constant, changing the injection speed, changing the relaxation time of the polypropylene melt, or scaling the mold will result in similar geometric shapes of the flow field. If the size of the mold is not changed, the geometric shape of the flow field will be the same. Since the dimensionless equation represents a similar system of all sizes, we verified the above conclusion through molecular dynamics simulations at a smaller scale. After further improvement of the micro simulation system, there is a possibility of visualizing the formation process of flow marks. This would greatly aid in the advancement of theory and the elimination of flow marks in production and experiments. This work also illustrates that the methodology of dimensional analysis plus molecular dynamics simulation may be applied to a wider range of other systems, scaling down large systems and thus significantly reducing their computational effort.

cond-mat.soft↗

arXiv2022

OpBoost: A Vertical Federated Tree Boosting Framework Based on Order-Preserving Desensitization

Vertical Federated Learning (FL) is a new paradigm that enables users with non-overlapping attributes of the same data samples to jointly train a model without directly sharing the raw data. Nevertheless, recent works show that it's still not sufficient to prevent privacy leakage from the training process or the trained model. This paper focuses on studying the privacy-preserving tree boosting algorithms under the vertical FL. The existing solutions based on cryptography involve heavy computation and communication overhead and are vulnerable to inference attacks. Although the solution based on Local Differential Privacy (LDP) addresses the above problems, it leads to the low accuracy of the trained model. This paper explores to improve the accuracy of the widely deployed tree boosting algorithms satisfying differential privacy under vertical FL. Specifically, we introduce a framework called OpBoost. Three order-preserving desensitization algorithms satisfying a variant of LDP called distance-based LDP (dLDP) are designed to desensitize the training data. In particular, we optimize the dLDP definition and study efficient sampling distributions to further improve the accuracy and efficiency of the proposed algorithms. The proposed algorithms provide a trade-off between the privacy of pairs with large distance and the utility of desensitized values. Comprehensive evaluations show that OpBoost has a better performance on prediction accuracy of trained models compared with existing LDP approaches on reasonable settings. Our code is open source.

cs.LG↗

arXiv2021

Lower and upper bounds of quantum battery power in multiple central spin systems

We study the energy transfer process in quantum battery systems consisting of multiple central spins and bath spins. Here with "quantum battery" we refer to the central spins, whereas the bath serves as the "charger". For the single central-spin battery, we analytically derive the time evolutions of the energy transfer and the charging power with arbitrary number of bath spins. For the case of multiple central spins in the battery, we find the scaling-law relation between the maximum power $P_{max}$ and the number of central spins $N_B$. It approximately satisfies a scaling law relation $P_{max}\propto N_{B}^α$, where scaling exponent $α$ varies with the bath spin number $N$ from the lower bound $α=1/2$ to the upper bound $α=3/2$. The lower and upper bounds correspond to the limits $N\to 1$ and $N\gg N_B$, respectively. In thermodynamic limit, by applying the Holstein-Primakoff (H-P) transformation, we rigorously prove that the upper bound is $P_{max}=0.72 B A \sqrt{N} N_{B}^{3/2}$, which shows the same advantage in scaling of a recent charging protocol based on the Tavis-Cummins model. Here $B$ and $A $ are the external magnetic field and coupling constant between the battery and the charger.

quant-ph↗

arXiv2019

Grüneisen parameters for Lieb-Liniger and Yang-Gaudin models

Using the Bethe ansatz solution, we analytically study expansionary, magnetic and interacting Grüneisen parameters (GPs) for one-dimensional (1D) Lieb-Liniger and Yang-Gaudin models. These different GPs elegantly quantify the dependences of characteristic energy scales of these quantum gases on the volume, the magnetic field and the interaction strength, revealing the caloric effects resulted from the variations of these potentials. The obtained GPs further confirm an identity which is incurred by the symmetry of the thermal potential. We also present universal scaling behavior of these GPs in the vicinities of the quantum critical points driven by different potentials. The divergence of the GPs not only provides an experimental identification of non-Fermi liquid nature at quantum criticality but also elegantly determine low temperature phases of the quantum gases. Moreover, the pairing and depairing features in the 1D attractive Fermi gases can be captured by the magnetic and interacting GPs, facilitating experimental observation of quantum phase transitions. Our results open to further study the interaction- and magnetic-field-driven quantum refrigeration and quantum heat engine in quantum gases of ultracold atoms.

cond-mat.quant-gas↗

arXiv2019

A Capsule Network for Recommendation and Explaining What You Like and Dislike

User reviews contain rich semantics towards the preference of users to features of items. Recently, many deep learning based solutions have been proposed by exploiting reviews for recommendation. The attention mechanism is mainly adopted in these works to identify words or aspects that are important for rating prediction. However, it is still hard to understand whether a user likes or dislikes an aspect of an item according to what viewpoint the user holds and to what extent, without examining the review details. Here, we consider a pair of a viewpoint held by a user and an aspect of an item as a logic unit. Reasoning a rating behavior by discovering the informative logic units from the reviews and resolving their corresponding sentiments could enable a better rating prediction with explanation. To this end, in this paper, we propose a capsule network based model for rating prediction with user reviews, named CARP. For each user-item pair, CARP is devised to extract the informative logic units from the reviews and infer their corresponding sentiments. The model firstly extracts the viewpoints and aspects from the user and item review documents respectively. Then we derive the representation of each logic unit based on its constituent viewpoint and aspect. A sentiment capsule architecture with a novel Routing by Bi-Agreement mechanism is proposed to identify the informative logic unit and the sentiment based representations in user-item level for rating prediction. Extensive experiments are conducted over seven real-world datasets with diverse characteristics. Our results demonstrate that the proposed CARP obtains substantial performance gain over recently proposed state-of-the-art models in terms of prediction accuracy. Further analysis shows that our model can successfully discover the interpretable reasons at a finer level of granularity.

cs.IR↗

arXiv2019

SECaps: A Sequence Enhanced Capsule Model for Charge Prediction

Automatic charge prediction aims to predict appropriate final charges according to the fact descriptions for a given criminal case. Automatic charge prediction plays a critical role in assisting judges and lawyers to improve the efficiency of legal decisions, and thus has received much attention. Nevertheless, most existing works on automatic charge prediction perform adequately on high-frequency charges but are not yet capable of predicting few-shot charges with limited cases. In this paper, we propose a Sequence Enhanced Capsule model, dubbed as SECaps model, to relieve this problem. Specifically, following the work of capsule networks, we propose the seq-caps layer, which considers sequence information and spatial information of legal texts simultaneously. Then we design a attention residual unit, which provides auxiliary information for charge prediction. In addition, our SECaps model introduces focal loss, which relieves the problem of imbalanced charges. Comparing the state-of-the-art methods, our SECaps model obtains 4.5% and 6.4% absolutely considerable improvements under Macro F1 in Criminal-S and Criminal-L respectively. The experimental results consistently demonstrate the superiorities and competitiveness of our proposed model.

cs.AI↗

arXiv2019

Knowledge Squeezed Adversarial Network Compression

Deep network compression has been achieved notable progress via knowledge distillation, where a teacher-student learning manner is adopted by using predetermined loss. Recently, more focuses have been transferred to employ the adversarial training to minimize the discrepancy between distributions of output from two networks. However, they always emphasize on result-oriented learning while neglecting the scheme of process-oriented learning, leading to the loss of rich information contained in the whole network pipeline. Inspired by the assumption that, the small network can not perfectly mimic a large one due to the huge gap of network scale, we propose a knowledge transfer method, involving effective intermediate supervision, under the adversarial training framework to learn the student network. To achieve powerful but highly compact intermediate information representation, the squeezed knowledge is realized by task-driven attention mechanism. Then, the transferred knowledge from teacher network could accommodate the size of student network. As a result, the proposed method integrates merits from both process-oriented and result-oriented learning. Extensive experimental results on three typical benchmark datasets, i.e., CIFAR-10, CIFAR-100, and ImageNet, demonstrate that our method achieves highly superior performances against other state-of-the-art methods.

cs.LG↗

arXiv2014

Expurgated Random-Coding Ensembles: Exponents, Refinements and Connections

This paper studies expurgated random-coding bounds and exponents for channel coding with a given (possibly suboptimal) decoding rule. Variations of Gallager's analysis are presented, yielding several asymptotic and non-asymptotic bounds on the error probability for an arbitrary codeword distribution. A simple non-asymptotic bound is shown to attain an exponent of Csiszár and Körner under constant-composition coding. Using Lagrange duality, this exponent is expressed in several forms, one of which is shown to permit a direct derivation via cost-constrained coding which extends to infinite and continuous alphabets. The method of type class enumeration is studied, and it is shown that this approach can yield improved exponents and better tightness guarantees for some codeword distributions. A generalization of this approach is shown to provide a multi-letter exponent which extends immediately to channels with memory. Finally, a refined analysis expurgated i.i.d. random coding is shown to yield a O\big(\frac{1}{\sqrt{n}}\big) prefactor, thus improving on the standard O(1) prefactor. Moreover, the implied constant is explicitly characterized.

cs.IT↗

arXiv2012

Observation of Multiphoton-induced Fluorescence from Nano Graphene Oxide and Its Applications in In vitro and In vivo Bioimaging

In the present paper, we observed both two-photon and three-photon induced distinct photoluminescence from GO nanoparticles under fs laser excitation. Conjugated with PEG molecules, GO nanoparticles exhibited high chemical stability, and could effectively label HeLa cells. Imaged with a two-photon scanning microscope, GO nanoparticles were observed to localize in the mitochondria, endoplasmic reticulum, Golgi and lysosome of HeLa cells. Furthermore, GO nanoparticles were micro-injected into the brain of a black mouse, and in vivo two-photon luminescence imaging illustrated that GO nanoparticles located at 300 μm depth in the brain could be clearly distinguished.

physics.optics↗

arXiv2010

Distortion Outage Probability in MIMO Block-Fading Channels

We study analogue source transmission over MIMO block-fading channels with receiver-only channel state information. Unlike previous work which considers the end-to-end expected distortion as a figure of merit, we study the distortion outage probability. We first consider the well known transmitter informed bound, which yields a benchmark lower bound to the distortion outage probability of any coding scheme. We next compare the results with source-channel separation. The key difference from the expected distortion approach is that if the channel code rate is chosen appropriately, source-channel separation can not only achieve the same diversity exponent, but also the same distortion outage probability as the transmitter informed lower bound.

cs.IT↗

arXiv2008

Quantitative Imaging of Single, Unstained Viruses with Coherent X-rays

Since Perutz, Kendrew and colleagues unveiled the structure of hemoglobin and myoglobin based on X-ray diffraction analysis in the 1950s, X-ray crystallography has become the primary methodology used to determine the 3D structure of macromolecules. However, biological specimens such as cells, organelles, viruses and many important macromolecules are difficult or impossible to crystallize, and hence their structures are not accessible by crystallography. Here we report, for the first time, the recording and reconstruction of X-ray diffraction patterns from single, unstained viruses. The structure of the viral capsid inside a virion was visualized. This work opens the door for quantitative X-ray imaging of a broad range of specimens from protein machineries, viruses and organelles to whole cells. Moreover, our experiment is directly transferable to the use of X-ray free electron lasers, and represents a major experimental milestone towards the X-ray imaging of single macromolecules.

physics.bio-ph↗