Home Page

Mario Annunziato is researcher in Mathematics, in the field of Numerical Analysis, at the "Università degli Studi di Salerno" in Italy since the year 2004.
His interests focuse to the numerical solution of Partial Differential Equations (PDE) and Integral Equations, related to stochastic processes and stochastic optimal control, and to modelling and applications of stochastic processes.

The goals of the research are:

to find the probability density function(PDF) of a stochastic processes by numerically solving PDEs of parabolic and hyperbolic type, or Volterra Integral equations, by ensuring that the discrete PDF be positive (or monotone) and conservative.
to develop numerical schemes for optimization problems with PDE constraints, related to the optimal control of stochastic processes.
to provide equation modelling for random phenomena.

Numerical methods for Piecewise Deterministic Processes

"Piecewise Deterministic Processes" (PDP) are a general model for stochastic point processes where a noise affects the motion of a state function, only at some random point epochs. They have applications to queue systems, reliability analysis and stochastic hybrid systems.
A stochastic processes can be almost completely described by its probability density (transition) function (PDF). The Chapman-Kolmogorov equation is an abstract equation for the PDF. In the case of PDP the CK equation can take the form of system of first order hyperbolic PDEs or Volterra integral equations . These equations sometime are named Liouville Master Equation or generalized Fokker-Planck equations.
The system of hyperbolic PDEs has initial Cauchy conditions and the evolutory PDF solution must be non-negative and total probability conservative in time. Further, depending on the type of point process, the PDEs can have special non-local boundary conditions, where the boundary conditions are not assigned functions, like classical Dirichelet or Neumann, but depend on the unknown PDF by integrals over the interior of the domain.
PDEs with this problem formulation are very little investigated. The development of stable, positive and mass preserving numerical methods is the aim of this research task.

Optimal control of stochastic processes

The possibility to control a stochastic process is a very interesting and stimulating research subject, and have potentially many applications in science, engineering and finance. In the optimal control theory the state of a system is controlled by minimizing (maximizing) an objective function of the state. Mathematically it is a constrained minimization problem.
For stochastic models, in the current scientific literature, the problem is formulated with an average of the cost functional of the stochastic state. The solution of this optimal control problem can be found by solving an Hamilton-Jacobi-Bellman equation.
In this research we propose to use the PDF as representative of the state of the system, define the objective as a functional of the PDF, and use the Fokker-Planck equation as constraint of the optimization problem. This is a new and unexplored framework in the field of optimization.
The solution can be found by formulating the minimization problem as an optimality system of PDEs, in order to find the reduced gradient of the objective and its vanishing point.
The aim of this research is to develop numerical methods to solve the optimality system and special minimization techniques for this non-linear optimization problem.

Mario Annunziato - Researcher -

Numerical methods for Piecewise Deterministic Processes

Optimal control of stochastic processes