稳健性(进化)
控制理论(社会学)
计算机科学
分段
数学优化
人工神经网络
趋同(经济学)
非线性系统
动态规划
滑模控制
收敛速度
功能(生物学)
理论(学习稳定性)
自适应控制
领域(数学)
最优控制
分段线性函数
鲁棒控制
非线性规划
控制工程
模式(计算机接口)
非线性控制
控制系统
线性规划
作者
Chao Jia,Xueqian Li,Xinyu Li
摘要
ABSTRACT In the paper, the adaptive dynamic programming (ADP) control method with prescribed‐time performance is investigated. Two design methods are proposed. First, a performance function associated with the prescribed‐time sliding mode surface is constructed, and a target function incorporating a prescribed‐time varying function is designed to derive a weight update law with prescribed‐time characteristics. This approach solves the nonlinear Hamilton–Jacobi–Bellman (HJB) equation and enables the system to achieve prescribed‐time convergence. Then, by introducing a prescribed‐time varying function into the ADP neural network target function and designing a new piecewise time‐varying function for the control law, a more universal approach to realize the prescribed‐time optimal control is presented, which does not require any additional design tools and only involves pure weight updates, adhering to the general design concept in the ADP field while ensuring the system maintains prescribed‐time convergence. The stability analysis for both methods is elaborated. Finally, simulation results demonstrate the superiority of the proposed methods in terms of prescribed‐time convergence time. Comparatively, the first method benefits from the robustness of sliding mode and features a fast convergence speed, while the second method can achieve prescribed‐time stability by relying only on the weight update rate of ADP itself, without the need to superimpose other control methods.
科研通智能强力驱动
Strongly Powered by AbleSci AI