We investigate three distinct methods of incorporating all-sky imager (ASI) images into deep learning (DL) irradiance nowcasting. The first method relies on a convolutional neural network (CNN) to extract features directly from raw RGB images. The second method uses state-of-the-art algorithms to engineer 2D feature maps informed by domain knowledge, e.g., cloud segmentation, the cloud motion vector, solar position, and cloud base height. These feature maps are then passed to a CNN to extract compound features. The final method relies on aggregating the engineered 2D feature maps into time-series input. Each of the three methods were then used as part of a DL model trained on a high-frequency, 29-day dataset to generate multi-horizon forecasts of global horizontal irradiance up to 15 minutes ahead. The models were then evaluated using root mean squared error and skill score on 7 selected days of data. Aggregated engineered ASI features as model input yielded superior forecasting performance, demonstrating that integration of ASI images into DL nowcasting models is possible without complex spatially-ordered DL-architectures and inputs, underscoring opportunities for alternative imag
We study the $L^p$ concentration problem for the Born--Jordan distribution in dimension $d>1$, thus extending the one-dimensional analysis in [Stra-Svela-Trapasso, J. Math. Pures Appl. (2026)]. We show that the existence of concentration optimizers depends on the exponent $p$ with a critical threshold at $p_*(d)= \frac{2d}{d-2}$ for $d\geq2$ (with the understanding that $p_*(2)=\infty$). In particular, for subcritical exponents $1\leq p<p_*(d)$ we prove that the supremum is finite and is attained, whereas for supercritical exponents $p>p_*(d)$ we show that the functional is unbounded. We also provide the complete solution in the (significantly more) challenging critical regime in dimension $d=2$.
We obtain large sieve type inequalities for the Rayleigh quotient of the restriction of phase space representations of higher rank operators, via an operator analogue of the short-time Fourier transform (STFT). The resulting bounds are referred to as `quantum large sieve inequalities'. On the shoulders of Donoho and Stark, we demonstrate that these inequalities guarantee the recovery of an operator whose phase-space information is missing or unobservable over a 'measure-sparse' region $Ω$, by solving an $L^{1}$-minimization program. This is an operator version of what is commonly known as `Logan's phenomenon'. Moreover, our results can be viewed as a deterministic, continuous variable version, on phase space, of `low-rank' matrix recovery, which itself can be regarded as an operator version of (finite-rank) compressive sensing. Our results depend on an abstract large sieve principle for operators with integrable STFT and on a non-commutative analogue of the local reproducing formula in rotationally invariant domains (first stated by Seip for the Fock space of entire functions). As an application, we obtain concentration estimates for Cohen's class distributions and the Husimi funct
We study nonlinear concentration problems for time-frequency distributions in the Cohen class. Using recent techniques from quantum harmonic analysis (QHA) we provide both positive and negative results, such as sufficient conditions for the existence of optimizers in terms of the ``window operator'' and explicit examples where the supremum is never attained. We also study the structural properties of window operators, in particular operators that yield weakly continuous concentration functionals and operators for which the nonlinear concentration problem admits an optimizer, also beyond the Heisenberg representation. We then consider generalizations to the study of concentration problems for phase space representations of operators. We consider generalized Husimi distributions via quantum convolution, and their optimization problem when optimizing over Hilbert--Schmidt and density operators. Lastly, we consider representations of operators on double phase space, in the spirit of quantum time-frequency analysis, and give a full solution in terms of the Weyl symbols.
Solving jigsaw puzzles has been extensively studied. While most existing models focus on solving either small-scale puzzles or puzzles with no gap between fragments, solving large-scale puzzles with gaps presents distinctive challenges in both image understanding and combinatorial optimization. To tackle these challenges, we propose a framework of Evolutionary Reinforcement Learning with Multi-head Puzzle Perception (ERL-MPP) to derive a better set of swapping actions for solving the puzzles. Specifically, to tackle the challenges of perceiving the puzzle with gaps, a Multi-head Puzzle Perception Network (MPPN) with a shared encoder is designed, where multiple puzzlet heads comprehensively perceive the local assembly status, and a discriminator head provides a global assessment of the puzzle. To explore the large swapping action space efficiently, an Evolutionary Reinforcement Learning (EvoRL) agent is designed, where an actor recommends a set of suitable swapping actions from a large action space based on the perceived puzzle status, a critic updates the actor using the estimated rewards and the puzzle status, and an evaluator coupled with evolutionary strategies evolves the actio
We propose MoRe-ERL, a framework that combines Episodic Reinforcement Learning (ERL) and residual learning, which refines preplanned reference trajectories into safe, feasible, and efficient task-specific trajectories. This framework is general enough to incorporate into arbitrary ERL methods and motion generators seamlessly. MoRe-ERL identifies trajectory segments requiring modification while preserving critical task-related maneuvers. Then it generates smooth residual adjustments using B-Spline-based movement primitives to ensure adaptability to dynamic task contexts and smoothness in trajectory refinement. Experimental results demonstrate that residual learning significantly outperforms training from scratch using ERL methods, achieving superior sample efficiency and task performance. Hardware evaluations further validate the framework, showing that policies trained in simulation can be directly deployed in real-world systems, exhibiting a minimal sim-to-real gap.
In the past decade, the fourth-generation light source based on the combination of Energy Recovery Linac (ERL) and Free-Electron Laser (FEL) using superconducting linear accelerators has garnered significant attention. It holds immense potential, particularly in generating high-power Extreme Ultraviolet (EUV) light sources. This article primarily focuses on the physical design of an injector for ERL-FEL, based on a Very High Frequency (VHF) electron gun with a charge of 100 pC. The beam energy is accelerated to 10 MeV using 3-cell superconducting cavity. The optimization of beam parameters is conducted through employment of BMad and ASTRA simulations, incorporating the concept of Merger optimization. The beam emittance is less than 0.6 mm mrad, and the peak current at the injector exit exceeds 18 A. We present a new method to evaluate the Longitudinal Space Charge (LSC) effects in merger sections, which can be readily applied in design work. Furthermore, we introduce a novel type of merger. The performance of this new merger is comparable to the previously known optimum, the zigzag merger, offering a potential alternative solution for injectors in ERLs.
The injector for ERL-FEL has been widely researched. Unlike traditional linacs, the bunch in the injector for ERLs requires additional deflection and matching section at lower energies. It makes the bunch more susceptible to the effects of the Space Charge. This will lead to a degradation in beam quality. In this paper, we comprehensively analyze the impact of space charge on ERL-injector and propose new design concepts to further maintain beam quality.
Daubechies-type theorems for localization operators are established in the multi-variate setting, where Hagedorn wavepackets are identified as the proper substitute of the Hermite functions. The class of Reinhardt domains is shown to be the natural class of masks that allow for a Daubechies-type result. Daubechies' classical theorem is a consequence of double orthogonality results for the short-time Fourier transform. We extend double orthogonality to the quantum setting and use it to establish Daubechies-type theorems for mixed-state localization operators, a key notion of quantum harmonic analysis. Lastly, we connect the results to Toeplitz operators on quantum Gabor spaces.
This work introduces Transformer-based Off-Policy Episodic Reinforcement Learning (TOP-ERL), a novel algorithm that enables off-policy updates in the ERL framework. In ERL, policies predict entire action trajectories over multiple time steps instead of single actions at every time step. These trajectories are typically parameterized by trajectory generators such as Movement Primitives (MP), allowing for smooth and efficient exploration over long horizons while capturing high-level temporal correlations. However, ERL methods are often constrained to on-policy frameworks due to the difficulty of evaluating state-action values for entire action sequences, limiting their sample efficiency and preventing the use of more efficient off-policy architectures. TOP-ERL addresses this shortcoming by segmenting long action sequences and estimating the state-action values for each segment using a transformer-based critic architecture alongside an n-step return estimation. These contributions result in efficient and stable training that is reflected in the empirical results conducted on sophisticated robot learning environments. TOP-ERL significantly outperforms state-of-the-art RL methods. Thoro
The Large Hadron electron Collider (LHeC) is a proposed future particle physics project colliding 60 GeV electrons from a six-pass recirculating energy-recovery linac (ERL) with 7 TeV protons stored in the LHC. The ERL technology allows for much higher beam current and, therefore, higher luminosity than a traditional linac. The high-current, high-energy electron beam can also be used to drive a free electron laser (FEL). In this contribution, we examine how the LHeC ERL can serve as a source of high-energy photons for studies in nuclear physics, high-energy physics, Axion detection, dark energy, and protein crystallography. In the first section, we discuss the performance of the LHeC-based FEL, operated in the SASE mode for generating photon pulses at wavelengths ranging from 200 keV to 600 keV. In the second section, we investigate photon production via Laser Compton scattering (LCS).
The accuracy of sensor fusion algorithms are limited by either the intrinsic sensor noise, or by the quality of time synchronization of the sensors. While the intrinsic sensor noise only depends on the respective sensors, the error induced by quality of, or lack of, synchronization depends on the dynamics of the vehicles and robotic system and the magnitude of time synchronization errors. To meet their sensor fusion requirements, system designers must consider both which sensor to use and also how to synchronize them. This paper presents the Syncline model, a simple visual model of how time synchronization affects the accuracy of sensor fusion for different mobile robot platform. The model can serve as a simple tool to determine which synchronization mechanisms should be used.
NASA's Hubble Space Telescope has captured a spectacular red, white, and blue view of one of the Milky Way's oldest star clusters to celebrate the nation's 250th anniversary。 Hidden within the ancient cluster are clues to how exploding stars helped transform the young universe into one capable of forming planets and, eventually, life
A strange "chirping" signal from a distant supernova has revealed the birth of a magnetar, confirming that these incredibly magnetic neutron stars can power the universe's brightest stellar explosions。 The discovery also marks the first time Einstein's general relativity has been used to explain the mechanics of a supernova
Top robotics researchers and founders explain how robot autonomy is evolving
A new study suggests Earth may have been sending tiny hitchhikers to Venus for billions of years。 Researchers found that asteroid impacts could launch microbes into space, where some might survive the journey and end up suspended in Venus' clouds。 If future missions detect life there, there's a surprising chance it didn't originate on Venus at all—
Ancient asteroid impacts may have done more than reshape Earth's surface—they could have helped spark life itself。 New computer models show the collisions created enormous underground hydrothermal systems by cracking the planet's crust and allowing hot water to flow through it。 These long-lasting, life-friendly environments may have covered much of