Abstract
We establish the existence of Nash equilibria in a large class of differential games. More specifically, we show that certain differential games are compact and better reply secure when the strategy space is equipped with the weak topology. Therefore, we can claim that such games must have Nash equilibria even when the payoffs of some of the players are discontinuous. On one hand, verifying better reply security with respect to a nonmetrizable weak topology is technically involved and requires a number of theoretical tools from functional analysis. On the other hand, our approach will allow us to generalize a number of current existence results regarding linear-quadratic differential games, games of evasion and pursuit, as well as some classic differential games in economics.
Similar content being viewed by others
Notes
When the set of pure strategies available to of player i is V i , a relaxed control σ i is a measurable function from [0,T] to the set of all probability measure on V i . Note just as pure strategies can be embedded within mixed strategies, a mixed strategy can considered as a constant relaxed control, and therefore the set of mixed strategies can be strictly embedded within the set of relaxed controls.
Deriving the standard HJB equations normally requires various “smoothness” conditions on the payoffs and the value functions of the players (see Sect. 4.2 in [4]).
Lemma 1 actually holds for any Banach space.
In other words, for every t, the map (x,y i ,y −i )↦f i (x,y i ,y −i ,t) is continuous, and for every (x,y i ,y −i ), the map t↦f i (x,y i ,y −i ,t) is measurable.
References
Basar T, Olsder G (1998) Dynamic non-cooperative game theory. Classics in applied mathematics, vol 23. SIAM, Philadelphia
Berkovitz L (1989) A survey of recent results in differential games. Working paper, Purdue University
Caputo M (2005) Foundations of dynamic economic analysis. Cambridge University Press, Cambridge
Dockner E et al. (2006) Differential games in economics and management science. Cambridge University Press, Cambridge
Elliott RJ, Kalton N (1972) Values in differential games. Bull Am Math Soc 78:427–431
Eisele T (1982) Nonexistence and nonuniqueness of open loop equilibria in linear-quadratic differential games. J Optim 37(4):443–468
Engwerda J (1998) On the open loop Nash equilibrium. J Econ Dyn Control 22:729–762
Ho YC et al (1965) Differential games and optimal Pursuit–Evasion strategies. Trans Autom Control 385–389
Kamien M, Schwartz N (1992) Dynamic optimization. Advanced textbooks in economics, vol 31. Elsevier, Amsterdam
Karray S, Zaccour G (2005) A differential game of advertising for national and store brands. In: Haurie A, Zaccour G (eds) Dynamic games: theory and applications. Chap 11
Olech C (1976) Weak lower semicontinuity of integral functionals. J Optim Theory Appl 19(1):3–16
Parthasarathy T, Raghavan E (1975) Existence of saddle points and Nash equilibrium point for differential games. SIAM J Control 977–980
Petrosjan L, Zenkevich N (1996) Game theory, series on optimization. World Scientific, Singapore
Reed M, Simon B (1980) Functional analysis, 2nd edn. Academic Press, San Diego
Reny P (1999) On the existence of pure and mixed equilibria in discontinuous games. Econometrica 67(5):1029–1056
Rockafellar RT (1975) Existence theorems for general control problems of Bolza and Lagrange. Adv Math 15:315–333
Warga J (1967) Functions of relaxed controls. SIAM J Control 6(4):628–641
Zhang P et al. (2008) On the value of two-person zero-sum linear quadratic differential games. In: Proceedings of 44th IEEE conference on decision and control, pp 714–718
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Lemma A.1
Let α(t), v n(t) and v(t) be functions in L 2([0,T];ℝn). Assume \(v^{n}\mathop{\rightarrow}\limits^{w} v\), and let \(s^{n}(t)=\int^{t}_{0} \alpha(\tau)v^{n}(\tau)\, d\tau\) and \(s(t)= \int^{t}_{0} \alpha (\tau)v(\tau)\, d\tau\). Then s n pointwise converge to s on [0,T].
Proof
Given \(v^{n}\mathop{\rightarrow}\limits^{w} v\) in L 2([0,T];ℝn), \(s^{n}(t)=\int^{t}_{0} \alpha(\tau)v^{n}(\tau)\, d\tau\) can expressed as
where δ [0,t] is the indicator function of the interval [0,t], i.e., δ [0,t](τ)=1 when τ∈[0,t] and δ [0,t](τ)=0, otherwise. Clearly, for a fixed t, the function α(τ)δ [0,t](τ) is in L 2([0,T];ℝn) and by the definition of weak convergence, we must have
and, therefore, s n(t)→s(t) for every t∈[0,T]. □
Rights and permissions
About this article
Cite this article
Bagh, A. Better Reply Security and Existence of Equilibria in Differential Games. Dyn Games Appl 3, 325–340 (2013). https://doi.org/10.1007/s13235-012-0061-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13235-012-0061-8