Solution of a backward stochastic differential equation

299 Views Asked by At

In the paper Adapted solutions of backward stochastic differential equations by Pardoux and Peng the authors prove in Lemma 2.1 that the stochastic equation $$ x(t) + \int_t^1 f(s) ds + \int_t^1 (g(s) + y(s)) dW(s) = X, \quad t \in [0,1] $$ has a unique solution. Here $W(t)$ is a $k$-dimensional Brownian motion with natural filtration $\mathcal F_t$, $f \in M^2(0,1;\mathbb{R}^d)$ and $g \in M^2(0,1;\mathbb{R}^{d+k})$ ($M^2$ is the class of $\mathcal F_t$- progressively measurable processes indexed by $t \in [0,1]$ which are square integrable with values in $\mathbb{R}^d$ and $\mathbb{R}^{d+k}$ respectively), and $X$ is $\mathcal F_1$ measurable. By solution to the equation we mean $(x(t),y(t))$ (with values in $\mathbb{R}^d \times \mathbb{R}^{d+k}$) which is $\mathcal F_t$-adapted.

The unique solution, constructed in the proof, is given by $$ x(t) = E(X - \int_t^1 f(s) ds|\mathcal F_t), \quad t \in [0,1], $$ and $$ y(t) = \overline y(t) - g(t), \quad t \in [0,1], $$ where $\overline y(t) \in M^2(0,1;\mathbb{R}^{d+k})$ satisfies $$ \int_0^t \overline y(s) dW(s) = E(X - \int_0^1 f(s) ds| \mathcal F_t) - x(0), \quad t \in [0,1]. $$

How can I show that such $(x(t),y(t))$ is a solution? How can I show that it is unique?

1

There are 1 best solutions below

2
On

Here is how I see it by the definition of $y$ you have $\forall t>0$ : $$X=x(t) + \int_t^1 f(s) ds + \int_t^1 (g(s) + y(s)) dW(s)$$ $$X=x(t) + \int_t^1 f(s) ds + \int_t^1 \bar{y}(s) dW(s) (1)$$ And the term $$\int_t^1 \bar{y}(s) dW(s)=\int_0^1 \bar{y}(s) dW(s)-\int_0^t \bar{y}(s) dW(s)$$
$$= E(X - \int_0^1 f(s) ds| \mathcal F_1) - x(0) - E(X - \int_0^1 f(s) ds| \mathcal F_t) + x(0)$$ $$= X - \int_0^1 f(s) ds +\int_0^t f(s) ds - E(X - \int_t^1 f(s) ds| \mathcal F_t) $$

becomes in the end : $$= X - \int_t^1 f(s) ds - x_t $$ (because by definition : $x_t=E(X - \int_t^1 f(s) ds| \mathcal F_t) $

Putting this expression back into (1) everything collapse which proves the claim unless mistaken.