Reinforcement fine-tuning (RFT) is essential for enhancing the reasoning and generalization capabilities of large language models, but its success heavily relies on the quality of the training data. While data selection has been extensively studied in supervised learning, its role in reinforcement learning, particularly during the RFT stage, remains largely underexplored. In this work, we introduce RFT-Inf, the first influence estimator designed for data in reinforcement learning. RFT-Inf quantifies the importance of each training example by measuring how its removal affects the final training reward, offering a direct estimate of its contribution to model learning. To ensure scalability, we propose a first-order approximation of the RFT-Inf score by backtracking through the optimization process and applying temporal differentiation to the sample-wise influence term, along with a first-order Taylor approximation to adjacent time steps. This yields a lightweight, gradient-based estimator that evaluates the alignment between an individual sample’s gradient and the average gradient direction of all training samples, where a higher degree of alignment implies greater training utility. Extensive experiments demonstrate that RFT-Inf consistently improves reward performance and accelerates convergence in reinforcement fine-tuning.

Encryption algorithms are usually applied to images in cloud environment to protect privacy, which will make image retrieval difficult. Existing encrypted image retrieval schemes mainly follow the two-step strategy, which firstly extract features from encrypted images and then use these features to conduct retrieval. However, the feature extraction procedure brings additional computational expense, and the extracted feature will result in privacy leakage. In this paper, we propose a new encrypted image retrieval scheme in end-to-end manner. End-to-end means we combine the feature extraction process and the retrieval process, which can reduce computational complexity. Specifically, images are encrypted by DC coefficients encryption, AC coefficients permutation and block permutation. After encryption, the encrypted images can be directly input into our retrieval model, which is based on Vision Transformer (Vit), without any process to conduct retrieval. What’s more, we propose a new type of data augmentation method named random block selection transformation (RST), which is compatible with the Vit backbone and can improve the model performance. Experiment result shows that our scheme can achieve superior retrieval accuracy than other state-to-the-art encrypted image retrieval schemes, while relatively high security can be retained.

In the DC microgrid cluster system, due to the large number of converters, there are many operation modes and switching frequencies. The traditional modeling methods are difficult to balance the accuracy of the model and the simplicity of calculation and are not suitable for different switching frequency systems. In view of the above problems, this paper uses simplified discrete time mapping model to model the system. It combines the state space average model with the discrete time mapping model, which greatly improves the simplicity and accuracy of modeling. Taking the photovoltaic-energy storage system as an example, this paper analyzes the nonlinear behavior of the system and predicts the critical control parameters when the Hopf bifurcation occurs in the system. The eigenvalue sensitivity analysis is used to determine the eigenvalue change rate and change trend when the control parameters change, which provides guidance for the selection of parameters in practical applications. Finally, the high precision of the model is verified by simulation, and the applicability and effectiveness of the method in different switching frequency systems are verified by experiments.

Benefited from harmonics well-defined and easy design of filter, the ON-OFF voltage-mode control strategy suitable for the very-high-frequency (VHF, 30~300MHz) dc-dc converter has been developed in recent year. However, the potential risk of this control strategy is that unreasonable parameter design will lead to system instability. Moreover, the conventional stability methods are also difficult to adopt in VHF converters due to its feature of high circuit order and many operating modes as well as extremely high switching frequency. Hence, this article proposes a new stability analysis method suitable for VHF converter employing the ON-OFF voltage-mode control strategy. Firstly, the combined characteristic of short-period switching signal and long-period modulation signal is considered, which dividing the VHF converter into short-period open-loop system and long-period closed-loop system. Then, by establishing both system discrete iterative map models, the effect of converter and controller parameters on stability are revealed by using bifurcation diagram and eigenvalue locus as well as Jacobian matrix. As a result, the established system models and stability analysis can obtain the stability boundary and predict the nonlinear behavior, which helps to guide the parameters design to ensure system stability. Finally, experimental results evaluated the correctness of the theoretical analysis.

To reduce switching losses of very-high-frequency (VHF) DC-DC converters, the quasi-resonance method is popular for achieving soft-switching. Yet, the widely adopted Class E and Class$\Phi _{2}$circuits may result in high voltage stress on switching devices, making devices selection difficult. Focusing on this issue, the zero-voltage-transition (ZVT) technique is introduced into VHF DC-DC converters and a novel ZVT clamped VHF DC-DC topology is proposed in this paper. It consists of two boost units. And the drive signals of both MOSFETs are complementary and have a duty cycle of 0.5. And by appropriate design considerations, the parasitic parameters of both MOSFETs and both diodes are absorbed in the circuit. Moreover, the resonance operation of this circuit only occurs before and after the switching action. And thanks to the clamping effect of the diodes, the voltage stress on all switching devices is significantly limited. Finally, an experimental prototype with operating frequency of 10 MHz, 9 V input and 21 V/20 W output, is built to verify the performances of the proposed converter, which has a higher measurement efficiency of 88%.

Dynamic performance analysis and controller design of wireless power transfer (WPT) systems have always been tough problems because the existing dynamic models of WPT systems have disadvantages, such as high order, incompleteness, and lack of intuition. In this article, a dynamic modeling method based on a singular perturbation theory for a high-order WPT system with a Cuk-based bridgeless rectifier at the receiving side is proposed, which overcomes the disadvantages of the dynamic models mentioned above. In terms of time-scale separation characteristics brought by singular perturbation, a reduced-order model can be derived from the original high-order WPT system model. The proposed model is able to completely present the dynamic behaviors of the whole WPT system precisely. Also, a closed-loop control design method based on the reduced-order model is presented, which greatly simplifies the design process. The experimental results show the great effectiveness of the proposed model for transient state observation, and the output voltage or current of the closed-loop WPT system remains stable despite significant disturbances in load or transfer distance.

Malicious attackers may disrupt the safety of autonomous systems through compromising sensors to feed wrong measurements to the controller. This article proposes attack-resilient sensor fusion that combines local sensor readings and shared sensing information from multiple sources. The method results in higher resilience against sensor attacks through jointly considering sensing noise and uncertain communication delay. To be specific, we first identify the considerable impact of the delay on determining attacked sensors. Second, we present a novel two-dimensional abstract sensor model, where each measurement is augmented as a probabilistic interval based on the convolution of the noise and delay. Third, we propose a fusion algorithm that admits the fused value with highest joint probability distribution of the intervals to tolerate corrupted measurements. Finally, we demonstrate the effectiveness of our method in a vehicle-platoon case study using extensive simulations and testbed experiments.

Summary In recent years, electric vehicles (EVs) receive intensive attention due to their environment‐friendly nature and outstanding energy efficiency. However, there are still obstacles in EVs' popularity. Battery related issues are most concerned from users' perspective. While the evolution in battery technology may help to address the aforementioned problems eventually, other solutions are in urgent demand for the time being. In this paper, we propose an innovative charge scheduling method that could significantly improve EV users' charging experience with the nowadays battery technology. Specifically, the proposed method takes into account the characteristics of the batteries to be charged as well as the constraints of the external power grid; thus, a plan can be devised for efficiently allocating the power of the charging station to the EVs through cyber‐physical system. Our method prioritizes jobs based on their marginal utilities to maximize user satisfaction with limited power resource. The power allocation in our method is managed globally and constrained by specific rules to avoid waste or overload in the power grid. Lastly, we simulate several power scheduling methods in charging stations, which validates that the user satisfaction under our method is higher than that of other traditional method.

The concept and technologies of fractional calculus have been increasingly applied to model and design circuits and electrical elements. In this work, by combining the principle of harmonic balance and equivalent small parameter method, we develop a new method of steady-state analysis for fractional-order DC-DC converters in discontinuous conduction mode (DCM). The effectiveness of the proposed scheme is confirmed by an example of fractional-order Boost converter in DCM. Both simulations and experimental results show that the periodic steady-state solutions of state variables obtained by the proposed scheme are accurate.

This paper proposes a novel quasi-Z source network with multiple switch-inductor (SL) cells and Cockroft-walton (CW) voltage multipliers. The proposed topology can further enhance the boost ability as the increasing of the count of the cascaded SL cells and CW cells. The utilization of CW cells at the output side can greatly reduce the voltage stress of the diodes and active switch, which makes the topology more suitable for dc-dc power conversion where higher gain and lower voltage stress are demanded. The performance of the topology is analyzed and a scale-down 180 W prototype is implemented to verify the theoretical analysis.

This paper proposes a transient modeling and analysis method for the fractional-order DC-DC converters in continuous conduction mode (CCM), in which a fractional-order Buck converter is taken as an example. By extending the usage of equivalent small parameter method (ESP), the converter is modeled by a general state vector differential equation. Then by adopting the principle of harmonic balance, fractional differential of state variables is converted into linear operations of exponential functions, and different harmonic are separated. Thus the main transient oscillation components of state variables can be obtained according to Grünwald-Letnikov's definition of fractional derivation, and the data are processed approximately by polynomial fitting of triangular functions. The results are introduced back to the model to obtain the values of the rest harmonic magnitudes. So the final solution of state variables is in the form of the summation of harmonic contents. To confirm the correctness of the results of the proposed method, waveforms of the analytical solution are put together with those from circuit simulation in PSIM. These results demonstrate the high effectiveness of the proposed method.

Based on the fact that both the phase and amplitude characteristics of fractional capacitor are related to its order, the introduction of fractional capacitor will make power factor correction (PFC) more flexible compared with the one using conventional capacitor. In order to verify the PFC function with fractional capacitor, an equivalent fractional capacitor model composed of an inverter and a resistor in series is presented in this paper. As the inverter is served as a controlled voltage source, the order of the proposed fractional capacitor can be varied between 0 and 2 easily just by changing the control parameters of the inverter, and the power level of the fractional capacitor model is the same as that of the inverter, which is able to be used in high power occasion. Finally, the simulation and experimental results are provided to validate the feasibility of power factor corrector with the proposed fractional capacitor model.

Negative-saliency PMSM in which Ldis higher than Lqcan achieve wider speed range, therefore it has a broad application prospect in electric vehicles. The parameters vary greatly under different operation conditions, leading to deteriorate the performance of the controller. In this paper, an adaptive control method based on online parameter identification is proposed. In constant torque region, adaptive maximum torque per ampere control is realized based on online parameter identification. In constant power region, adaptive field weakening control based on a PI regulator is used. Simulation and experiment results verify the reasonableness and correctness of the proposed control strategies.

A new method of steady-state analysis for non-integer order DC/DC converters operating in continuous-conduction-mode (CCM) is put forward. First, the periodic and time-variant non-integer order DC/DC converter system is modeled by a nonlinear vector differential equation. Then by combining the principle of harmonic balance and equivalent small parameter method, the vector differential equation is solved, and analytical periodic steady-state solutions are obtained. The analytical results on a non-integer order Boost converter agree well with those obtained by modeling simulations using MATLAB software.

Imaging mass spectrometry is a method for understanding the molecular distribution in a two-dimensional sample. This method is effective for a wide range of molecules, but generates a large amount of data. It is difficult to extract important information from these large datasets manually and automated methods for discovering important spatial and spectral features are needed. Independent component analysis and non-negative matrix factorization are explained and explored as tools for identifying underlying factors in the data. These techniques are compared and contrasted with principle component analysis, the more standard analysis tool. Independent component analysis and non-negative matrix factorization are found to be more effective analysis methods. A mouse cerebellum dataset is used for testing.

RENATI is an acronym for recontextualizing narratives for tangible interfaces. It serves as an umbrella term for our art/research experiments within a hybrid environment that uses oral narratives, and non-generative and immersive art with sensing technologies to create tangible narratives. In this paper we introduce our first prototype, which uses a custom-built mannequin to allow viewers to engage with a multi-viewpoint story titled Flying Over Purgatory.

Imaging mass spectrometry can be used to reveal spatial distributions of multiple molecular species in a 2D biological sample. Due to the large amount of data produced by this technology, it is difficult and time-consuming to manually extract meaningful results from imaging mass spectrometry experimentation. We have developed and implemented an original approach to easily and consistently process mass spectrometry imaging data with the goal of automatically identifying interesting regions of molecule expression. Based on multivariate analysis techniques such as principal component analysis, the system allows researchers to conveniently define and visualize spatial regions based on spectral similarity. Features of our system are demonstrated on mouse cerebellum data.