Pruning error analysis for a class of curse-of-dimensionality free methods

In the context of computational nonlinear optimal control, curse-of-dimensionality (CoD) refers to the phenomenon of exponential growth of computational cost with respect to the dimension of state and input space. It is well-known that CoD is the major drawback of grid-based computational methods, which are consequently restricted their applications to low dimensional problems. Switching linear quadratic regulators (SLQR) is a class of nonlinear optimal control problems for which a CoD free method has been developed. However, it has been observed that this CoD free method suffers from a different form of computational complexity known as curse-of-complexity (CoC) which refers to the phenomenon that the number of quadratics necessary to represent the value functions of SLQR problems increases exponentially with respect to time horizon. Pruning is the key method of tackling this class of complexity at the cost of introducing pruning errors. This paper develops a framework that can be used to analyse pruning errors.

[1]  Peter M. Dower,et al.  A max-plus based fundamental solution for a class of discrete time linear regulator problems ☆ , 2013, 1306.5060.

[2]  Dimitri P. Bertsekas,et al.  Dynamic Programming and Optimal Control, Two Volume Set , 1995 .

[3]  J. Quadrat.  Numerical methods for stochastic control problems in continuous time , 1994 .

[4]  Sean R Eddy,et al.  What is dynamic programming? , 2004, Nature Biotechnology.

[5]  Jianghai Hu,et al.  On the Value Functions of the Discrete-Time Switched LQR Problem , 2009, IEEE Transactions on Automatic Control.

[6]  Bo Lincoln,et al.  Efficient pruning of search trees in LQR control of switched linear systems , 2000, Proceedings of the 39th IEEE Conference on Decision and Control (Cat. No.00CH37187).

[7]  William M. McEneaney,et al.  A new fundamental solution for differential Riccati equations arising in control , 2008, Autom..

[8]  William M. McEneaney,et al.  A Curse-of-Dimensionality-Free Numerical Method for Solution of Certain HJB PDEs , 2007, SIAM J. Control. Optim..

[9]  W.M. McEneaney,et al.  Curse-of-complexity attenuation in the curse-of-dimensionality-free method for HJB PDEs , 2008, 2008 American Control Conference.

[10]  William M. McEneaney,et al.  Max-plus methods for nonlinear control and estimation , 2005 .

[11]  Huan Zhang,et al.  A pruning algorithm for managing complexity in the solution of a class of linear non-quadratic regulator problems , 2013, 2013 Australian Control Conference.

[12]  William M. McEneaney,et al.  Convergence Rate for a Curse-of-Dimensionality-Free Method for a Class of HJB PDEs , 2009, SIAM J. Control. Optim..

[13]  William M. McEneaney,et al.  Curse of dimensionality reduction in max-plus based approximation methods: Theoretical estimates and improved pruning algorithms , 2011, IEEE Conference on Decision and Control and European Control Conference.

[14]  Zheng Qu,et al.  Contraction of Riccati flows applied to the convergence analysis of the max-plus curse of dimensionality free method , 2013, 2013 European Control Conference (ECC).

[15]  B. Anderson,et al.  Optimal control: linear quadratic methods , 1990 .