Methods of regularization in nonsmooth problems of dynamic optimization (Q1304132): Difference between revisions

It is considered the extremal problem \[ \dot{x}\in F(x),\quad t\in I=[t_{1},t_{2}],\tag{1} \] \[ p\in P,\quad p=(x_{1},x_{2}),\quad x_{1}=x(t_{1}), \quad x_{2}=x(t_{2}),\tag{2} \] \[ x(t)\in G\quad \forall t\in I,\tag{3} \] \[ J(p)\to \min,\tag{4} \] where \(x\in R^{n}\), \(F\) is a locally Lipschitzian multivalued mapping with nonempty compact values, \(P\) and \(G\) are nonempty closed subsets of \(R^{2n}\) and \(R^{n}\), respectively, \(J\) is a locally Lipschitzian function. The problem (1)--(4) is searched for a minimum in the class of all locally Lipschitzian vector-valued functions \(x\) defined on its own time interval \(I\). The main difficulties in the study of the problem (1)--(4) arise due to the nonsmoothness of its data, the multivaluedness of the right-hand side of the differential inclusion (1), and the state constraints (3). In this paper are considered different methods of approximations for nonsmooth problems of dynamic optimization. These methods are based on the reduction of the initial nonsmooth problem to a family of smooth ones. Then the initial problem is reduced to the classical optimal control problem without state and endpoint constraints.