Wyniki wyszukiwania - BazTech

1

Convergence of finite-dimensional approximations for mixed-integer optimization with differential equations

Hante Falk M., Schmidt Martin

Control and Cybernetics

|

2019

|

Vol. 48, No. 2

209--226

EN

We consider a direct approach to solving the mixedinteger nonlinear optimization problems with constraints depending on initial and terminal conditions of an ordinary differential equation. In order to obtain a finite-dimensional problem, the dynamics are approximated using discretization methods. In the framework of general one-step methods, we provide sufficient conditions for the convergence of this approach in the sense of the corresponding optimal values. The results are obtained by considering the discretized problem as a parametric mixed-integer nonlinear optimization problem in finite dimensions, where the step size for discretization of the dynamics is the parameter. In this setting, we prove the continuity of the optimal value function under a stability assumption for the integer feasible set and second-order conditions from nonlinear optimization. We address the necessity of the conditions on the example of pipe sizing problems for gas networks.

2

Online learning algorithm for zero-sum games with integral reinforcement learning

Vamvoudakis K. G., Vrabie D., Lewis F. L.

Journal of Artificial Intelligence and Soft Computing Research

|

2011

|

Vol. 1, No. 4

315--332

EN

In this paper we introduce an online algorithm that uses integral reinforcement knowledge for learning the continuous-time zero sum game solution for nonlinear systems with infinite horizon costs and partial knowledge of the system dynamics. This algorithm is a data based approach to the solution of the Hamilton-Jacobi-Isaacs equation and it does not require explicit knowledge on the system’s drift dynamics. A novel adaptive control algorithm is given that is based on policy iteration and implemented using an actor/ disturbance/critic structure having three adaptive approximator structures. All three approximation networks are adapted simultaneously. A persistence of excitation condition is required to guarantee convergence of the critic to the actual optimal value function. Novel adaptive control tuning algorithms are given for critic, disturbance and actor networks. The convergence to the Nash solution of the game is proven, and stability of the system is also guaranteed. Simulation examples support the theoretical result.

3

Time-parametric control : uniform convergence of the optimal value functions of discretized problems

Gugat M.

Control and Cybernetics

|

1999

|

Vol. 28, no 1

7-33

EN

The problem of time-optimal control of linear hyperbolic systems is equivalent to the computation of the root of the optimal value function of a time-parametric program, whose feasible set is described by a countable system of moment equations. To compute this root, discretized problems with a finite number of equality constraints can be used. In this paper, we show that on a certain time-interval, the optimal value functions of the discretized problems converge uniformly to the optimal value function of the original problem. We also give sufficient conditions fot Lipschitz and Hoelder continuity of the optimal value function of the original problem.