Approximation of solutions to the optimal control problems for systems with maximum (Q2284394)

This paper describes a method for the approximate solution of optimal control problems, where the value of the input depends on the maximal values attained by the control parameter and the phase vector: \[ \dot{x}(t) = f(t,x(t),x_m(t)) + A(x)\xi(t,u(t),u_m(t)), \] where \(u\) is the control parameter, and \(x_m\) and \(u_m\) represent the supremum values attained by the functions \(x\) and \(u\) in an interval \([g(t),\gamma(t)]\). The solution proposed is to use an averaging method, namely to solve the problem \[ \dot{y}(t) = \bar{f}(y(t),y_m(t)) + A(y(t))v(t), \] where \(v\) is the new control parameter and the function \(\bar{f}\) is defined as \[ \frac{1}{T}\int_0^T f(t,x)dt. \] The authors propose an algorithm to find a correspondence between the averaged control \(v\) and the control \(u\), based on discretising the interval \([0,T]\) into subintervals of predefined length \(T_0\) and show that under suitable assumptions of continuity and the existence of relevant limits, there is a correspondence between the controls \(u\) and \(v\). Furthermore, the objective values of the averaged system and the original system can be made arbitrarily close.