Systems & Control Letters 60 (2011) 344–349 Contents lists available at ScienceDirect Systems & Control Letters journal homepage: www.elsevier.com/locate/sysconle Existence of optimal controls for systems driven by FBSDEs Khaled Bahlali a , Boulekhrass Gherbal b , Brahim Mezerdi b, a IMATH, UFR Sciences, USTV, B.P. 20132, 83957 La Garde, Cedex, France b Laboratory of Applied Mathematics, University Med Khider, BP 145, Biskra 07000, Algeria article info Article history: Received 19 April 2010 Received in revised form 29 January 2011 Accepted 28 February 2011 Available online 29 March 2011 Keywords: Forward backward stochastic differential equation Stochastic control Weak convergence Existence abstract We prove the existence of optimal relaxed controls as well as strict optimal controls for systems governed by non linear forward–backward stochastic differential equations (FBSDEs). Our approach is based on weak convergence techniques for the associated FBSDEs in the Jakubowski S-topology and a suitable Skorokhod representation theorem. © 2011 Elsevier B.V. All rights reserved. 1. Introduction In this paper, we study the existence of optimal controls for systems driven by FBSDEs of the form X t = x + t 0 b(s, X s , U s )ds + t 0 σ(s, X s )dW s , Y t = g (X T ) + T t f (s, X s , Y s , U s )ds T t Z s dW s (M T M t ) (1.1) where b, σ , f and g are given functions, (W t , t 0) is a standard Brownian motion, defined on some filtered probability space (Ω, F , F t , P ), satisfying the usual conditions. X , Y , Z are square integrable adapted processes and M is a square integrable martingale which is orthogonal to W . The control variable U t , called strict control, is a measurable, F t -adapted process with values in a compact metric space A. The expected cost on the time interval [0, T ] is of the form J (U ) = E [ l(Y 0 ) + T 0 h(t , X t , Y t , U t )dt ] . (1.2) Partially supported by CMEP, PHC Tassili no. 07 MDU 705 and Marie Curie ITN no. 213841-2. Corresponding author. Tel.: +213 33 74 77 88; fax: +213 33 74 77 88. E-mail addresses: bahlali@univ-tln.fr (K. Bahlali), bgherbal@yahoo.fr (B. Gherbal), bmezerdi@yahoo.fr (B. Mezerdi). The objective of the controller is to minimize this cost function, over the class U of admissible controls, that is, adapted processes with values in some set A, called the action space. A control u is called optimal if it satisfies J ( u) = inf{J (u), u U}. In the case of forward Itô’s SDEs, the existence of such a strict optimal control follows from the convexity of the image of the action space A by the mapping (b(t , x, .), σ 2 (t , x, .), h(t , x, .)), which is known as the Roxin-type convexity condition, see for instance [1–3]. Without this convexity condition, an optimal control may fail to exist in U. It should be noted that the set U is not equipped with a compact topology. The idea is then to introduce a new class R of admissible controls, in which the controller chooses at time t , a probability measure q t (da) on the control set A, rather than an element u t A. These are called relaxed controls. Using compactification techniques, Fleming [4] derived the first existence result of an optimal relaxed control for SDEs with uncontrolled diffusion coefficient. The case of SDEs with a controlled diffusion coefficient has been solved by El-Karoui et al. [1], where the optimal relaxed control is shown to be Markovian. Linear backward stochastic differential equations (BSDEs) have been studied in the early seventies by Bismut [5], in connection with the stochastic version of the Pontriagin maximum principle. More precisely, the adjoint process in the maximum principle satisfies a linear BSDE. The first existence and uniqueness result for non linear BSDEs has been proved by Pardoux and Peng [6]. This important paper has given rise to a huge literature on BSDEs and has become a powerful tool in many fields such as financial mathematics, optimal control, stochastic games, semi linear and quasi linear partial differential equations, differential geometry 0167-6911/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.sysconle.2011.02.011