ID 原文 译文
56358 经逐步分析得出,控制判别器的Lipschitz常数是解决GANs不稳定性问题的关键,进而提出一种有针对性的梯度惩罚技术来解决此问题. Unlike other penalty techniques, we penalize the Lipschitz constant of the discriminator, which isthe key to dealing with the instability problem (this amounts to controlling the Lipschitz constant of the dis?criminator).
56359 最后,本文从损失函数的振荡幅度(收敛性)、梯度总体变化趋势,以及网络整体性能3个方面进行了全面对比实验. We performed a series of experimental comparisons from three different perspectives: the oscillationamplitude of the loss function (convergence), the general variation trend of the gradient, and the holistic perfor?mance of the network.
56360 结果显示,本文所提出的惩罚技术对处理GANs训练的不稳定性问题具有显著的效果. The results demonstrated that the proposed technique has a significant and positive effecton the training instability in GANs.
56361 针对一类具有传感器故障和不对称输入死区的非线性多输入多输出非严格反馈系统,本文提出一种自适应神经网络容错控制方案. In this paper, an adaptive neural network fault-tolerant control (FTC) strategy is proposed for a classof multi-input and multi-output (MIMO) nonlinear systems in nonstrict-feedback form, in which sensor faultsand nonsymmetric dead zone inputs are considered.
56362 控制器的设计以反步法为框架,采用自适应神经网络控制方法处理传感器故障,利用死区斜率的有界性补偿输入死区对系统性能造成的影响,同时引入动态面控制技术克服"计算爆炸"的问题. Based on the backstepping method, the desired controlleris designed. The adaptive neural network control approach is proposed to handle the problem of sensor faults. The effect of dead zone inputs on the system performance can be compensated by virtue of the boundedness ofdead zone slopes.
56363 该控制方法不仅能够保证闭环系统中所有信号半全局一致最终有界,而且能使跟踪误差收敛至原点附近的紧集内. Furthermore, the dynamic surface control technique is introduced to prevent an “explosion ofcomplexity”. The proposed control method ensures that all the signals of the closed-loop system are semi-globallyuniformly ultimately bounded, and that the tracking errors converge to a compact set around the origin.
56364 最后通过两个仿真实验验证该控制方法的有效性. Finally,simulation results are provided to demonstrate the effectiveness of the proposed approach.
56365 针对一类具有未知干扰和多约束的不确定纯反馈非线性系统,提出一种基于干扰观测器的鲁棒自适应抗干扰控制方案. To solve the control problem for a class of uncertain pure feedback nonlinear systems subjectedto external disturbances and multiple constraints, an adaptive robust control methodology is proposed basedon a disturbance observer.
56366 该方法首先基于Butterworth低通滤波器和径向基神经网络设计非线性干扰观测器以实现对系统未知非线性函数和复合扰动的在线精确逼近,并消除"代数环"问题. To handle unknown nonlinearity and external unknown disturbances, a nonlineardisturbance observer is constructed based on a radial basis function neural network, which uses a Butterworthlow-pass filter to remove the algebraic loop problem.
56367 其次,为确保系统在状态受限、预设性能和输入饱和等多重约束的综合影响下能够对期望轨迹进行稳定跟踪,构造了一种新型的障碍Lyapunov函数,结合辅助有界函数、Nussbaum函数和一阶滑模微分器设计Backstepping控制器,并通过Lyapunov稳定理论分析闭环系统稳定性. Then, to guarantee that the system can stably track thedesired trajectory under the state constraints, input saturation, and prescribed tracking performance constraints,we developed a novel barrier Lyapunov function and a backstepping controller that combines an auxiliary boundedfunction, a Nussbaum function, and a first-order sliding-mode differentiator. Subsequently, the stability of theclosed-loop system is rigorously proved by Lyapunov analysis.