GAC, savings, and unbounded inputs

TL;DR

This paper extends control theory results by demonstrating that the existence of a Minimum Restraint Function guarantees global controllability and savings, even with unbounded controls, and analyzes polynomial control dynamics for simplification.

Contribution

It introduces the concept of a Minimum Restraint Function for unbounded controls and explores polynomial control systems, broadening the scope of controllability and savings guarantees.

Findings

Minimum Restraint Function guarantees GAC and savings with unbounded controls.

Control vector fields composed of Lipschitz maps, polynomials, and exponentials meet mild conditions.

Polynomial dynamics can be simplified via affine representability or subsystem reduction.

Abstract

Let a control system and a target be given on an open subset of an Euclidean space. The existence of a Control Lyapunov Function - namely a positive definite, semiconcave, solution of the Hamilton-Jacobi inequality corresponding to the control vector field -- guarantees Global Asymptotic Controllability (GAC). In this case, however, minimization is not an issue. Instead, if a Lagrangean with non-negative values is considered as well, an optimal control problem can be defined in relation to the corresponding integral functional. In the first part of the present paper we show that the existence of a Minimum Restraint Function -- a solution of a strict Hamilton-Jacobi inequality involving the Lagrangian and a non-negative "savings multiplier" -- provides not only global asymptotic controllability but also savings, namely a state-dependent upper bound for the infima. This extends a former result, where the control set was assumed to be compact. Here we allow unbounded controls and replace inputs' values' compactness with a quite mild hypothesis concerning the dependence of the data on inputs: such condition is met, for instance, by control vector fields that are compositions of Lipschitz maps with polynomials and exponentials of the control variable. In the second part of the paper we focus on the case when the dynamics is a polynomial in the control variable. Through some analysis of convexity properties of vector-valued polynomials' ranges, we prove some simplified versions of the main result, in terms of either affine representability or reduction to weak subsystems for the original dynamics.