Uncertain saddle point equilibrium differential games with non-anticipating strategies
Introduction
In the past few decades, game theory has been an active research field in operations research and control theory. Von Meumann and Morgenstern [24] first established the modern game theory. Later, Isaacs [15] studied a two person zero sum differential game model in a dynamical system which initiated the research of differential game theory. Pontryagin [25] considered a class of differential games with the maximum principle theorem. Friedman [12] and Berkovitz [2] investigated the existence theorem and approximation method for differential games.
In previous work, one important way to solve the two person zero sum differential game is to transform the original problem to solving a PDE (called HJI equaiton). An important assumption is that the value function of the game is assumed to be sufficiently smooth (e.g. twice differentiable) to make sense of the related HJI equation. Nevertheless, this assumption is usually impossible to be achieved. Many researchers [5], [9], [11] has worked on this difficulty with some relaxed assumptions. The breakthrough is the establishment of concept of non-anticipating strategy and viscosity solution (see e.g. [6], [8], [10], [18]).
Stochastic differential game [1], [3], [7], [14], [28] also received much attention. However, noises in some particular dynamical systems do not behave like randomness, such as the price of new stock, bridge strength and oil field reserves. There are no enough samples to ensure the estimated probability distribution of the noises is close enough to the long-run cumulative frequency. Hence, stochastic differential equations are not able to appropriately model these dynamical systems [23]. To estimate this kind of indeterministic noises, people have to invite some domain experts to evaluate the belief degree that each event may occur. Liu [19] founded uncertainty theory in 2007 to rationally deal with personal belief degrees. Nowadays, uncertain theory has been applied to many fields (see e.g. [4], [26], [27], [30], [34], [35]). Thus, for those dynamical systems which cannot be appropriately described by stochastic differential equations, we may use uncertain differential equations. Two person zero sum uncertain differential games were analyzed (see e.g. [13], [29], [31], [32]).
In this paper, we consider a two person zero sum uncertain differential game with non-anticipating strategies. The rest of this paper is organized as follows. In Section 2, we review some basic concepts about uncertainty theory. In Section 3, we formulate our saddle point equilibrium game model and introduce the non-anticipating strategy. In Section 4, we discuss the properties of the value function. In the Section 5, we establish the relationship between the value function and the uncertain HJI equation with the viscosity solution. In the last section, we give an example to illustrate our results.
Section snippets
Preliminaries
In this section, we introduce some important basic concepts about uncertainty theory, which are used throughout the paper.
Uncertainty theory is a branch of axiomatic mathematics to deal with human uncertainty arising from the belief degrees. Let Γ be a nonempty set and be a σ-algebra over Γ. Each element is called an event. Then uncertain measure is used to evaluate the belief degree that each event Λ may occur. The axiomatic definition is as follows.
Definition 1 [19] A set function defined on
Problem formulation
In this section, we formulate our uncertain saddle point equilibrium differential game model and provide some estimates about the system states. In addition, we introduce the notion of non-anticipating strategy.
We consider a control system described by an uncertain differential equation as follows: where with being some non-empty closed convex sets. In the above system, Xs is the state
Properties of the Elliott–Kalton lower and upper value
In this section, we discuss some basic properties of the Elliott–Kalton upper and lower value functions. First, we provide some preparations.
Lemma 2 Let ξ and η be two uncertain variables, and α ∈ (0, 1]. If ξ(γ) ≤ η(γ), ∀γ ∈ Γ, we have . Proof Since ξ(γ) ≤ η(γ) for any γ ∈ Γ, we have Thus, we have which yields
Then we know that i.e. . □ Lemma 3 Let Ct be a Liu process. And K
Viscosity solutions to uncertain Hamilton–Jacobi–Isaacs equations
In this section, we prove that the Elliott–Kalton upper and lower value functions are unique viscosity solutions to uncertain Hamilton–Jacobi–Isaacs equations. First, we introduce the concept of viscosity solution.
Definition 7 Assume is continuous. Also is assumed to be continuous. A continuous function J(t, x) is called a viscosity solution to the following Hamilton–Jacobi equation
provided that Eq. (10) holds with and for any
An example
Now, we provide an example to illustrate the significance of the viscosity solution approach for the uncertain saddle point differential game. HJI equation is an important indirect way to study the property of value function of a game. However the HJI equation may have no classical solution.
Consider an uncertain system: where the admissible set is . And the cost functional is given by For any
Conclusion
In this paper, we introduced the concept of non-anticipating strategy with which we define the Elliott–Kalton value function for uncertain saddle point equilibrium differential game. The continuity of this value function was investigated. Furthermore, we proved that the Elliott–Kalton value function satisfies the dynamic programming condition. Finally, we utilized the viscosity solution theory to characterize the value function with an uncertain Hamilton–Jacobi–Isaacs equation.
Acknowledgments
This work is supported by the National Natural Science Foundation of China (No. 61673011).
References (35)
The cauchy problem for a nonlinear first order partial differential equation
J. Differ. Equ.
(1969)- et al.
Zero-sum stochastic differential games and backward equations
Sys. Control Lett.
(1995) - et al.
Stability and optimal control for uncertain continuous-time singular systems
Eur. J. Control
(2017) - et al.
On a class of linear stochastic differential games
IEEE Trans. Automat. Contr.
(1968) The existence of value and saddle point in games of fixed duration
SIAM J. Control Optim.
(1985)- et al.
Stochastic differential games and viscosity solutions of Hamilton–Jacobi–Bellman–Isaacs equations
SIAM J. Control Optim.
(2008) - et al.
Indefinite LQ optimal control with equality constraint for discrete-time uncertain systems
Jpn. J. Ind. Appl. Math.
(2016) - et al.
Hamilton’s theory and generalized solutions of the Hamilton–Jacobi equation
J. Math. Mech.
(1964) - et al.
Viscosity solutions of Hamilton–Jacobi equations
Trans. Am. Math. Soc.
(1983) The existence of value in stochastic differential games
SIAM J. Control Optim.
(1976)
The existence of value in differential games
Mem. Am. Math. Soc.
On solving certain nonlinear partial differential equations by accretive operator methods
Israel J. Math.
Differential games and representation formulas for solutions of Hamilton–Jacobi–Isaacs equations
Indiana U. Math. J.
Differential Games
Uncertain Shapley value of coalitional game with application to supply chain alliance
Appl. Soft Comput.
Differential Games
Uniqueness of unbounded viscosity solutions of Hamilton–Jacobi equations
Indiana U. Math. J.
Cited by (6)
Uncertain stochastic hybrid zero-sum games based on forward uncertain difference equations and backward stochastic difference equations
2024, Journal of Computational and Applied MathematicsSome results about uncertain differential equations with time-dependent delay
2020, Applied Mathematics and ComputationCitation Excerpt :Considering the fluctuation of financial markets from time to time, Wang and Ning [21] also designed another uncertain currency model with floating interest rates. In addition, uncertain differential equations have also been applied to optimal control [4,9,15], differential game [18,23] and so on. Until now, uncertain differential equations have been successfully extended in many directions, including uncertain differential equations with jumps [26], backward uncertain differential equations [6], uncertain delay differential equations [5] and so on.
Linear quadratic zero-sum game for time-delayed uncertain stochastic systems
2024, Optimal Control Applications and MethodsUncertain differential game with Hurwicz criterion
2023, Journal of Ambient Intelligence and Humanized ComputingTwo-person cooperative uncertain differential game with transferable payoffs
2021, Fuzzy Optimization and Decision Making