A simplification of the backpropagation-through-time algorithm for optimal neurocontrol
A simplification of the backpropagation-through-time algorithm for optimal neurocontrol
复制标题
用于最佳神经控制的反向传播时间算法的简化
DOI：
10.1109/72.557698
复制
发表时间：
1997
期刊：
影响因子：
--
通讯作者：
V. Gorrini
中科院分区：
文献类型：
--
作者：
H. Bersini;V. Gorrini
时间反向传播（BPTT）是反向传播的时间扩展，它允许多层神经网络逼近最优状态反馈控制律，前提是该过程的某些先验知识（雅可比矩阵）可用。本文提出了一种简化的BPTT算法，它更接近于动态规划的最优性原则。除了更简单，新的算法是耗时更少，并允许在某些情况下发现更好的控制律。一个正式的理由，这种简化是试图通过混合拉格朗日微积分BPTT与贝尔曼-汉密尔顿-雅可比方程。由于这种简化的改进说明了两个最优控制问题：会合和生物反应器。
Backpropagation-through-time (BPTT) is the temporal extension of backpropagation which allows a multilayer neural network to approximate an optimal state-feedback control law provided some prior knowledge (Jacobian matrices) of the process is available. In this paper, a simplified version of the BPTT algorithm is proposed which more closely respects the principle of optimality of dynamic programming. Besides being simpler, the new algorithm is less time-consuming and allows in some cases the discovery of better control laws. A formal justification of this simplification is attempted by mixing the Lagrangian calculus underlying BPTT with Bellman-Hamilton-Jacobi equations. The improvements due to this simplification are illustrated by two optimal control problems: the rendezvous and the bioreactor.