Overview
Research introduces a control framework applied to stochastic compartmental models within epidemiology. This framework diverges from direct control of the stochastic system, instead focusing on optimal control of an associated Fokker-Planck equation. The objective is to guide the distribution of potential solutions from the stochastic system towards a predefined desirable state. This methodology is designed to facilitate robust control mechanisms, addressing uncertainties present in both the system's dynamics and its initial conditions.
Research Context
The study addresses the challenge of controlling epidemics modeled stochastically. Traditional approaches to epidemic control often contend with inherent uncertainties. The proposed framework specifically targets these uncertainties by shifting the control focus from the stochastic system itself to its probabilistic evolution described by a Fokker-Planck equation. This allows for a more distributed, rather than point-wise, control strategy over the possible states the system might occupy.
Approach
The methodology centers on formulating and analyzing a partial differential equation (PDE) constrained optimization problem. This problem incorporates the Fokker-Planck equation as a constraint. A key aspect of the approach is the proof of existence for optimal controls, which is established through an analysis of the control-to-state map. Furthermore, the optimal controls are characterized using the Pontryagin minimum principle.
For implementation and numerical approximation of optimal control maps, the researchers describe the application of the sequential quadratic Hamiltonian method. This numerical technique assists in translating the theoretical framework into actionable computations.
Findings
- The research developed a control framework for stochastic compartmental models in epidemiology.
- This framework employs optimal control of an associated Fokker-Planck equation rather than directly controlling the stochastic system.
- The goal of this control is to steer the distribution of possible solutions of the stochastic system to a desired state.
- The framework facilitates robust control mechanisms, accounting for uncertainty in dynamics and initial data.
- A partial differential equation constrained optimization problem was formulated and analyzed.
- Proof of existence of optimal controls was provided through analysis of the control-to-state map.
- Optimal controls were characterized via the Pontryagin minimum principle.
- The sequential quadratic Hamiltonian method is applicable for numerical approximations of optimal control maps within this framework.
- The method was demonstrated using a minimal stochastic susceptible-infected-recovered (SIR) model.
- Different choices of cost functionals were used in the SIR model demonstration, representing varied policy-maker concerns.
Potential Applications
The method was demonstrated using a minimal stochastic susceptible-infected-recovered (SIR) model. This demonstration included the exploration of different cost functionals, which are designed to represent varying concerns of policy-makers. This suggests that the framework can be adapted to align with diverse policy objectives in epidemic management.