arXiv subjects

Pu Zhang

Publications and source records attributed to Pu Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2023

Heterogeneous Trajectory Forecasting via Risk and Scene Graph Learning

Heterogeneous trajectory forecasting is critical for intelligent transportation systems, but it is challenging because of the difficulty of modeling the complex interaction relations among the heterogeneous road agents as well as their agent-environment constraints. In this work, we propose a risk and scene graph learning method for trajectory forecasting of heterogeneous road agents, which consists of a Heterogeneous Risk Graph (HRG) and a Hierarchical Scene Graph (HSG) from the aspects of agent category and their movable semantic regions. HRG groups each kind of road agent and calculates their interaction adjacency matrix based on an effective collision risk metric. HSG of the driving scene is modeled by inferring the relationship between road agents and road semantic layout aligned by the road scene grammar. Based on this formulation, we can obtain effective trajectory forecasting in driving situations, and superior performance to other state-of-the-art approaches is demonstrated by exhaustive experiments on the nuScenes, ApolloScape, and Argoverse datasets.

cs.CV↗

arXiv2023

Enhancing Mapless Trajectory Prediction through Knowledge Distillation

Scene information plays a crucial role in trajectory forecasting systems for autonomous driving by providing semantic clues and constraints on potential future paths of traffic agents. Prevalent trajectory prediction techniques often take high-definition maps (HD maps) as part of the inputs to provide scene knowledge. Although HD maps offer accurate road information, they may suffer from the high cost of annotation or restrictions of law that limits their widespread use. Therefore, those methods are still expected to generate reliable prediction results in mapless scenarios. In this paper, we tackle the problem of improving the consistency of multi-modal prediction trajectories and the real road topology when map information is unavailable during the test phase. Specifically, we achieve this by training a map-based prediction teacher network on the annotated samples and transferring the knowledge to a student mapless prediction network using a two-fold knowledge distillation framework. Our solution is generalizable for common trajectory prediction networks and does not bring extra computation burden. Experimental results show that our method stably improves prediction performance in mapless mode on many widely used state-of-the-art trajectory prediction baselines, compensating for the gaps caused by the absence of HD maps. Qualitative visualization results demonstrate that our approach helps infer unseen map information.

cs.CV↗

arXiv2023

FEND: A Future Enhanced Distribution-Aware Contrastive Learning Framework for Long-tail Trajectory Prediction

Predicting the future trajectories of the traffic agents is a gordian technique in autonomous driving. However, trajectory prediction suffers from data imbalance in the prevalent datasets, and the tailed data is often more complicated and safety-critical. In this paper, we focus on dealing with the long-tail phenomenon in trajectory prediction. Previous methods dealing with long-tail data did not take into account the variety of motion patterns in the tailed data. In this paper, we put forward a future enhanced contrastive learning framework to recognize tail trajectory patterns and form a feature space with separate pattern clusters. Furthermore, a distribution aware hyper predictor is brought up to better utilize the shaped feature space. Our method is a model-agnostic framework and can be plugged into many well-known baselines. Experimental results show that our framework outperforms the state-of-the-art long-tail prediction method on tailed samples by 9.5% on ADE and 8.5% on FDE, while maintaining or slightly improving the averaged performance. Our method also surpasses many long-tail techniques on trajectory prediction task.

cs.CV↗

arXiv2022

MFV approach to robust estimate of neutron lifetime

Aiming at evaluating the lifetime of the neutron, we introduce a novel statistical method to analyse the updated compilation of precise measurements including the 2022 dataset of Particle Data Group (PDG). Based on the minimization for the information loss principle, unlike the median statistics method, we apply the most frequent value (MFV) procedure to estimate the neutron lifetime, irrespective of the Gaussian or non-Gaussian distributions. Providing a more robust way, the calculated result of the MFV is $τ_n=881.16^{+2.25}_{-2.35}$ s with statistical bootstrap errors, while the result of median statistics is $τ_n=881.5^{+5.5}_{-3}$ s according to the binomial distribution. Using the different central estimates, we also construct the error distributions of neutron lifetime measurements and find the non-Gaussianity, which is still meaningful.

physics.data-an↗

arXiv2022

Monic modules and semi-Gorenstein-projective modules

The category ${\rm gp}(Λ)$ of Gorenstein-projective modules over tensor algebra $Λ= A\otimes_kB$ can be described as the monomorphism category ${\rm mon}(B, {\rm gp}(A))$ of $B$ over ${\rm gp}(A)$. In particular, Gorenstein-projective $Λ$-modules are monic. In this paper, we find the similar relation between semi-Gorenstein-projective $Λ$-modules and $A$-modules, via monic modules, namely, ${\rm mon}(B, \ ^\perp A) = {\rm mon}(B, A)\cap \ ^\perp Λ.$ Using this, it is proved that if $A$ is weakly Gorenstein, then $Λ$ is weakly Gorenstein if and only each semi-Gorenstein-projective $Λ$-modules are monic; and that if $B = kQ$ with $Q$ a finite acyclic quiver, then $Λ$ is weakly Gorenstein if and only if $A$ is weakly Gorenstein. However, this relation itself does not answer the question whether there exist double semi-Gorenstein-projective $Λ$-modules which are not monic. Using the recent discovered examples of double semi-Gorenstein-projective $A$-modules which are not torsionless, we positively answer this question, by explicitly constructing a class of double semi-Gorenstein-projective $T_2(A)$-modules with one parameter such that they are not monic, and hence not torsionless. The corresponding results are obtained also for the monic modules and semi-Gorenstein-projective modules over the triangular matrix algebras given by bimodules.

math.RT↗

arXiv2022

Gorenstein-projective modules over short local algebras

Following the well-established terminology in commutative algebra, any (not necessarily commutative) finite-dimensional local algebra $A$ with radical $J$ will be said to be short provided $J^3 = 0$. As in the commutative case, we show: if a short local algebra $A$ has an indecomposable non-projective Gorenstein-projective module $M$, then either $A$ is self-injective (so that all modules are Gorenstein-projective) and then $|J^2| \le 1$, or else $|J^2| = |J/J^2| - 1$ and $|JM| = |J^2||M/JM|$. More generally, we focus the attention to semi-Gorenstein-projective and $\infty$-torsionfree modules, even to $\mho$-paths of length 2, 3 and 4. In particular, we show that the existence of a non-projective reflexive module implies that $|J^2| < |J/J^2|$ and further restrictions. In addition, we consider exact complexes of projective modules with a non-projective image. Again, as in the commutative case, we see that if such a complex exists, then $A$ is self-injective or satisfies the condition $|J^2| = |J/J^2| - 1.$ Also, we show that any non-projective semi-Gorenstein-projective module $M$ satisfies $Ext^1(M,M) \neq 0$. In this way, we prove the Auslander-Reiten conjecture (one of the classical homological conjectures) for arbitrary short local algebras. Many arguments used in the commutative case actually work in general, but there are interesting differences and some of our results may be new also in the commutative case.

math.RT↗

arXiv2022

Calibrating quantum hydrodynamic model for noble metals in nanoplasmonics

Quantum hydrodynamic model (QHDM) has become a versatile and efficient tool for studying plasmonics at the nanoscopic length scale. Yet its application to noble metals has not been sufficiently justified, in particular for situations where the metallic structures interface with dielectric material and electrons spill over the interfaces. In a recent work, we developed a refined QHDM, where the near-field effects and static polarization of metal ion lattice, and the electron affinity and static permittivity of the dielectric are incorporated. Here we perform a careful calibration of the model parameters for the refined QHDM. The model parameters are determined by benchmarking with (time-dependent) density functional theory calculations for special cases of simple metal. The predictive power of the refined QHDM with calibrated model parameters is faithfully demonstrated by the calculations of the optical responses from gold nanomatryoshkas of different sizes. The refined QHDM approach allows the quasinormal mode analysis for revealing the intrinsic optical properties of the nanoscopic metallic structures. We expect the well-calibrated refined QHDM would provide the nanoplasmonics community with a useful tool.

physics.optics↗

arXiv2022

Trajectory Forecasting from Detection with Uncertainty-Aware Motion Encoding

Trajectory forecasting is critical for autonomous platforms to make safe planning and actions. Currently, most trajectory forecasting methods assume that object trajectories have been extracted and directly develop trajectory predictors based on the ground truth trajectories. However, this assumption does not hold in practical situations. Trajectories obtained from object detection and tracking are inevitably noisy, which could cause serious forecasting errors to predictors built on ground truth trajectories. In this paper, we propose a trajectory predictor directly based on detection results without relying on explicitly formed trajectories. Different from the traditional methods which encode the motion cue of an agent based on its clearly defined trajectory, we extract the motion information only based on the affinity cues among detection results, in which an affinity-aware state update mechanism is designed to take the uncertainty of association into account. In addition, considering that there could be multiple plausible matching candidates, we aggregate the states of them. This design relaxes the undesirable effect of noisy trajectory obtained from data association. Extensive ablation experiments validate the effectiveness of our method and its generalization ability on different detectors. Cross-comparison to other forecasting schemes further proves the superiority of our method. Code will be released upon acceptance.

cs.CV↗

arXiv2021

A general framework of canonical quasinormal mode analysis for extreme nano-optics

Optical phenomena associated with extremely localized field should be understood with considerations of nonlocal and quantum effects, which pose a hurdle to conceptualize the physics with a picture of eigenmodes. Here we first propose a generalized Lorentz model to describe general nonlocal media under linear mean-field approximation and formulate source-free Maxwell's equations as a linear eigenvalue problem to define the quasinormal modes. Then we introduce an orthonormalization scheme for the modes and establish a canonical quasinormal mode framework for general nonlocal media. Explicit formalisms for metals described by quantum hydrodynamic model and polar dielectrics with nonlocal response are exemplified. The framework enables for the first time direct modal analysis of mode transition in the quantum tunneling regime and provides physical insights beyond usual far-field spectroscopic analysis. Applied to nonlocal polar dielectrics, the framework also unveils the important roles of longitudinal phonon polaritons in optical response.

physics.optics↗

arXiv2021

A Nonparametric Bayesian Framework for Uncertainty Quantification in Stochastic Simulation

When we use simulation to assess the performance of stochastic systems, the input models used to drive simulation experiments are often estimated from finite real-world data. There exist both input model and simulation estimation uncertainties in the system performance estimates. Without strong prior information on the input models and the system mean response surface, in this paper, we propose a Bayesian nonparametric framework to quantify the impact from both sources of uncertainty. Specifically, since the real-world data often represent the variability caused by various latent sources of uncertainty, Dirichlet Processes Mixtures (DPM) based nonparametric input models are introduced to model a mixture of heterogeneous distributions, which can faithfully capture the important features of real-world data, such as multi-modality and skewness. Bayesian posteriors of flexible input models characterize the input model estimation uncertainty, which automatically accounts for both model selection and parameter value uncertainty. Then, input model estimation uncertainty is propagated to outputs by using direct simulation. Thus, under very general conditions, our framework delivers an empirical credible interval accounting for both input and simulation uncertainties. A variance decomposition is further developed to quantify the relative contributions from both sources of uncertainty. Our approach is supported by rigorous theoretical and empirical study.

stat.ME↗

arXiv2021

Quasinormal mode theory for nanoscale electromagnetism with quantum surface responses

We report a self-consistent quasinormal mode theory for nanometer scale electromagnetism where the possible nonlocal and quantum effects are treated through quantum surface responses. With Feibelman's frequency-dependent \textit{d} parameters to describe the quantum surface responses, we formulate the source-free Maxwell's equations into a generalized linear eigenvalue problem to define the quasinormal modes. We then construct an orthonormal relation for the modes and consequently unlock the powerful toolbox of modal analysis. The orthonormal relation is validated by the reconstruction of the full numerical results through modal contributions. Significant changes in the landscape of the modes are observed due to the incorporation of the quantum surface responses for a number of nanostructures. Our semi-analytical modal analysis enables transparent physical interpretation of the spontaneous emission enhancement of a dipolar emitter as well as the near-field and far-field responses of planewave excitations in the nanostructures.

physics.optics↗

arXiv2020

Bright Optical Eigenmode with about 1 nm $^3$ Volume

Concentrating optical field in an eigenmode with a tiny volume is vitally important for light-matter interactions at the fundamental level and underpins new technologies. In the past decades, researchers have investigated various approaches to shrink light and so far managed to reduce the volume of optical eigenmodes, under the quantum-optical definition, down to several ten nm3. Here we report on the discovery, characterization and engineering of a class of extremely localized eigenmodes that are resonances of atomistic protrusions on a metallic host nanoparticle and feature quantum-optical mode volumes of below 1 nm3. We theoretically demonstrate that these extremely localized modes can be made bright with radiation efficiencies reaching 30% and provide up to 4x10^7 times intensity enhancement. The existence of bright eigenmodes with the volume comparable to a photon emitter foresees exciting new optical physics, such as ultrastrong coupling with single optical emitters, angstrom-resolution optical imaging, and atomic-scale single-molecule photochemistry.

physics.optics↗

arXiv2020

The Truncated Metallo-dielectric Omnidirectional Reflector: Collecting Single Photons in the Fundamental Gaussian Mode with 95% Efficiency

We propose a novel antenna structure which funnels single photons from a single emitter with unprecedented efficiency into a low-divergence fundamental Gaussian mode. Our device relies on the concept of creating an omnidirectional photonic bandgap to inhibit unwanted large-angle emission and to enhance small-angle defect-guided-mode emission. The new photon collection strategy is intuitively illustrated, rigorously verified and optimized by implementing an efficient body-of-revolution finite-difference time-domain method for in-plane dipole emitters. We investigate a few antenna designs to cover various boundary conditions posed by fabrication processes or material restrictions and theoretically demonstrate that collection efficiencies into the fundamental Gaussian mode exceeding 95% are achievable. Our antennas are broadband, insensitive to fabrication imperfections and compatible with a variety of solid-state emitters such as organic molecules, quantum dots and defect centers in diamond. Unidirectional and low-divergence Gaussian-mode emission from a single emitter may enable the realization of a variety of photonic quantum computer architectures as well as highly efficient light-matter interfaces.

physics.optics↗

arXiv2020

Multiscale Modeling and Analysis for High-fidelity Interferometric Scattering Microscopy

Interferometric scattering microscopy (iSCAT), as an ultrasensitive fluorescence-free imaging modality, has recently gain enormous attention and been rapidly developing from demonstration of principle to quantitative sensing. Here we report on a theoretical and experimental study for iSCAT with samples having structural dimensions that differ by 4-5 orders of magnitude. In particular, we demonstrate and intuitively explain the profound effects of sub-nanometer surface roughness of a glass coverslip and of a mica surface on the absolute signal and the shape of the point spread function of a gold nanoparticle. These quantities significantly affect the accuracies for determining the target size and position in all three dimensions. Moreover, we investigate a sample system mimicking a gold nanoparticle in a simplified cell environment and show position-dependent and even asymmetric point spread function of the nanoparticle. The multiscale study will facilitate the development of high fidelity iSCAT in real applications.

physics.optics↗

arXiv2020

On modules M such that M and M* are semi-Gorenstein-projective

Let A be an artin algebra. An A-module M is semi-Gorenstein-projective provided that Ext^i(M,A) = 0 for all i > 0. If M is Gorenstein-projective, then both M and its A-dual M* are semi-Gorenstein projective. As we have shown recently, the converse is not true, thus answering a question raised by Avramov and Martsinkovsky. The aim of the present note is to analyse in detail the modules M such that both M and M* are semi-Gorenstein-projective.

math.RT↗

arXiv2020

Koszul modules (and the $Ω$-growth of modules) over short local algebras

Following the well-established terminology in commutative algebra, any (not necessarily commutative) finite-dimensional local algebra $A$ with radical $J$ will be said to be short provided $J^3 = 0$. As in the commutative case, also in general, the asymptotic behavior of the Betti numbers of modules seems to be of interest. As we will see, there are only few possibilities for the growth of the Betti numbers of modules. We generalize results which are known for commutative algebras, but some of our results seem to be new also in the commutative case.

math.RT↗

arXiv2020

Characterization of Lipschitz Functions via Commutators of Multilinear Singular Integral Operators in Variable Lebesgue Spaces

In this paper, the main aim is to consider the boundedness of commutators of multilinear Calderón-Zygmund operators with Lipschitz functions in the context of the variable exponent Lebesgue spaces. Furthermore, the variable versions of the Lipschitz spaces are also discussed, and with the help of a key tool of a pointwise estimate involving the sharp maximal operator of the multilinear fractional commutator and certain associated maximal operators.

math.CA↗

arXiv2020

Temperature and Rate Dependent Constitutive Behaviors of Low Melt Field's Metal

Low melting point metals such as Field's metal and gallium are increasingly used as transition phases in stiffness-tuning soft materials and devices. Nevertheless, there is a knowledge gap on the fundamental constitutive behaviors of metals with melting points below 100 degree. This letter reports the stress-strain relationships of Field's metal at various temperatures and strain rates. This metal has the lowest melting point among metals whose constitutive behaviors have been studied systematically. Experimental results indicate that Field's metal exhibits a strain-softening behavior, in oppose to the strain-hardening behaviors of most engineering metals and alloys. A modified Johnson-Cook model is devised in this letter to describe the constitutive behaviors of Field's metal. The proposed metal will facilitate the design and analysis of functional materials and structures consisting of Field's metal.

cond-mat.mtrl-sci↗