Abstract:For a particular class of planar dynamics that are linear with respect to the control variable, we show that the feedback strategy ''null-singular null'' is minimizing the maximum of a coordinate over infinite horizon, under a L 1 budget constraint on the control. Moreover, we characterize the optimal cost as a function of the budget. The proof is based on an unusual use of the clock form. This result generalizes the one obtained formerly for the SIR epidemiological model to more general Kolmogorov dynamics, that we illustrate on other biological models.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to minimize the maximum value of a certain coordinate in an infinite - time range under the \(L^1\) constraint of the control input in a specific class of planar dynamic systems. Specifically, the paper focuses on dynamic systems that are linearly dependent on control variables and proposes a "Null - Singular - Null" (NSN) feedback strategy to achieve this goal.
### Background and Problem Description
In many practical problems, especially in biological and epidemiological models, we often need to optimize certain performance indicators under limited resources or budgets. For example, in epidemic control, we may hope to minimize the maximum number of infected people under limited social isolation measures. Such problems can usually be formalized as optimal control problems, where the objective function is the maximum value of a certain state variable, and the control input is subject to the \(L^1\) constraint.
### Specific Problems
The specific problems considered in the paper are:
- **Dynamic System**: A class of planar dynamic systems, whose state equations can be expressed as:
\[
\begin{cases}
\dot{x} = f_1(x, y)+g_1(x, y)u \\
\dot{y} = f_2(x, y)+g_2(x, y)u
\end{cases}
\]
where \(f_1, f_2, g_1, g_2\) are smooth functions of the state variables \(x, y\), and \(u\) is the control input.
- **Objective**: Minimize the maximum value of \(y\) in an infinite - time range when the control input \(u\) satisfies the \(L^1\) constraint:
\[
\inf_{u(\cdot)\in U}\sup_{t\geq0}y(t)
\]
where \(U\) represents the set of all measurable functions whose values are in \([0, 1]\) and satisfy the \(L^1\) constraint:
\[
\int_0^{+\infty}u(t)\,dt\leq K
\]
### Solution
The paper proposes a "Null - Singular - Null" (NSN) feedback control strategy, which is divided into three stages:
1. **Rapid Ascent Stage**: Raise \(y\) to the optimal peak as quickly as possible.
2. **Maintenance Stage**: Apply control to keep the peak of \(y\) unchanged until the budget is exhausted (singular arc).
3. **Release Stage**: When entering a region of the state space, the peak of \(y\) will not increase even without applying control.
### Main Results
- **Optimality Proof**: The paper proves that under certain conditions, the NSN strategy is optimal by using Green's Theorem and an unconventional clock form technique.
- **Application**: The paper also discusses the application of this result in some specific biological models, including the SIR epidemic model and the resource - consumer model.
### Conclusion
The paper provides a general framework for solving the minimization of the maximum value problem under the \(L^1\) constraint, and demonstrates the effectiveness of this framework through specific examples. This provides a theoretical basis for optimizing the performance of biological systems under resource - constrained conditions.