arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2401.13014eess.SYcs.SY
收藏

一种用于非线性连续时间H∞控制问题的新型策略迭代算法

A Novel Policy Iteration Algorithm for Nonlinear Continuous-Time H$\infty$ Control Problem

Qi Wang

AI总结:

针对非线性连续时间H∞控制中HJI方程难以求解的问题,提出α策略迭代(α-PI)强化学习方法,通过广义贝尔曼方程迭代,无需系统动力学知识,并验证了其有效性。

摘要:

非线性连续时间系统的H∞控制依赖于Hamilton-Jacobi-Isaacs(HJI)方程的解,由于HJI方程的非线性特性,已证明无法获得其闭式解。为了求解HJI方程,人们提出了许多迭代算法,且当在Banach空间中构造不动点方程时,大多数算法本质上都是牛顿法。牛顿法是一种局部优化方法,其收敛区域较小,需要初始猜测与解足够接近。而阻尼牛顿法增强了对初始条件的鲁棒性,并具有更大的收敛区域。本文提出了一种名为α策略迭代(α-PI)的新型强化学习方法,用于求解HJI方程。首先,通过构造阻尼牛顿迭代算子方程,得到广义贝尔曼方程(GBE)。GBE是贝尔曼方程的推广。然后,通过对GBE进行迭代,提出了一种无需系统内部动态知识的on-policy α-PI强化学习方法。第三,基于on-policy α-PI强化学习方法,我们开发了一种无需任何系统动态知识的off-policy α-PI强化学习方法。最后,分别推导了基于神经网络的on-policy和off-policy α-PI算法的自适应评判实现方案,并使用批量最小二乘法计算神经网络的权重参数。通过计算机仿真验证了off-policy α-PI算法的有效性。

英文摘要:

H{\infty} control of nonlinear continuous-time system depends on the solution of the Hamilton-Jacobi-Isaacs (HJI) equation, which has been proved impossible to obtain a closed-form solution due to the nonlinearity of HJI equation. In order to solve HJI equation, many iterative algorithms were proposed, and most of the algorithms were essentially Newton method when the fixed-point equation was constructed in a Banach space. Newton method is a local optimization method, it has small convergence region and needs the initial guess to be sufficiently close to the solution. Whereas damped Newton method enhances the robustness with respect to initial condition and has larger convergence region. In this paper, a novel reinforcement learning method which is named α-policy iteration (α-PI) is introduced for solving HJI equation. First, by constructing a damped Newton iteration operator equation, a generalized Bellman equation (GBE) is obtained. The GBE is an extension of bellman equation. And then, by iterating on the GBE, an on-policy α-PI reinforcement learning method without using knowledge regarding to the system internal dynamics is proposed. Third, based on the on-policy α-PI reinforcement learning method, we develop an off-policy α-PI reinforcement learning method without requiring any knowledge of the system dynamics. Finally, the neural-network based adaptive critic implementation schemes of on-policy and off-policy α-PI algorithms are derived respectively, and the batch least-squares method is used for calculating the weight parameters of neural networks. The effectiveness of the off-policy α-PI algorithm is verified through computer simulation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑