ID 原文 译文
52967 以此方法建立直线电机车辆的机电动力学模型，简单有效地解决了相关的机电动力学问题。 The electromechanical dynamics model of the linear motor vehicle is established by this method, and the related electromechanical dynamics problems are solved simply and effectively.
52968 针对存在参数摄动、负载扰动和未建模动态等不确定因素的永磁直线同步电机(PMLSM)位移系统，提出了一种基于固定时间干扰观测器的动态面反步控制方法。 For the permanent magnet linear synchronous motor(PMLSM) with uncertain factors such as parameter perturbation, load disturbance, unmodeled dynamics, and so on, a dynamic surface backstepping control method is proposed based on the fixed-time disturbance observer.
52969 首先，通过构造干扰观测器对系统的非匹配不确定项和匹配不确定项进行观测估计，且其观测误差能够在固定时间内收敛于0; Firstly, the disturbance observers are developed to observe the mismatched uncertainty and matched uncertainties of the system, and the observation errors can converge to 0 in fixed time.
52970 其次，将反步控制与动态面控制相结合完成PMLSM固定时间位移跟踪控制器的设计， Secondly, the fixed-time displacement tracking controllers for PMLSM are presented by combining the backstepping control with dynamic surface control,
52971 有效避免了常规反步控制存在的“微分爆炸”问题，并简化了系统控制器的设计过程。 which avoids the 'differential explosion' problem during using the conventional backstepping control, and simplifies the design processes of the system controllers.
52972 理论分析表明，PMLSM位移系统在本文所提控制方法的作用下能够在固定时间内收敛。 Theoretical analysis shows that the PMLSM displacement system can converge in fixed time under the action of the proposed control method.
52973 最后，与其他控制方法进行了仿真对比研究，仿真结果验证了所提方法的有效性。 Finally, the simulation results verify the effectiveness of the proposed control method through comparing with other control methods.
52974 本文从多任务系统以及算法硬件协同设计的角度出发，提出了一种面向多任务处理的神经网络加速器架构。 An accelerator architecture from the perspective of multi-task systems and algorithm-and-hardware co-design is proposed.
52975 在算法方面，提出了 一套多任务处理算法，能够有效挖掘多任务间的特征重复性和计算重用性; First, a simultaneous multi-task algorithm is proposed to exploit the repetitive feature and the computational reuse among the tasks.
52976 在架构设计方面，设计了能够利用网络参数共享实现多任务间计算重用的加速器架构。 Then a compact accelerator that fully utilizes the proposed algorithm is designed to process various vision tasks simultaneously.