Abstract and Applied Analysis

Abstract and Applied Analysis / 2011 / Article

Research Article | Open Access

Volume 2011 |Article ID 310910 | https://doi.org/10.1155/2011/310910

Guangchen Wang, Zhen Wu, "Mean-Variance Hedging and Forward-Backward Stochastic Differential Filtering Equations", Abstract and Applied Analysis, vol. 2011, Article ID 310910, 20 pages, 2011. https://doi.org/10.1155/2011/310910

Mean-Variance Hedging and Forward-Backward Stochastic Differential Filtering Equations

Academic Editor: Gabriel Turinici
Received22 May 2011
Accepted28 Jun 2011
Published20 Sep 2011


This paper is concerned with a mean-variance hedging problem with partial information, where the initial endowment of an agent may be a decision and the contingent claim is a random variable. This problem is explicitly solved by studying a linear-quadratic optimal control problem with non-Markov control systems and partial information. Then, we use the result as well as filtering to solve some examples in stochastic control and finance. Also, we establish backward and forward-backward stochastic differential filtering equations which are different from the classical filtering theory introduced by Liptser and Shiryayev (1977), Xiong (2008), and so forth.

1. Introduction and Problem Formulation

We begin with a finite time horizon [0,๐‘‡] for ๐‘‡>0, a complete filtered probability space (ฮฉ,โ„ฑ,(โ„ฑ๐‘ก),โ„™) on which an ๐‘๐‘š-valued standard Brownian motion (๐‘Š(โ‹…)) is defined. Moreover, we let the natural filtration โ„ฑ๐‘ก=๐œŽ{๐‘Š(๐‘ );0โ‰ค๐‘ โ‰ค๐‘ก}, 0โ‰ค๐‘กโ‰ค๐‘‡, and โ„ฑ=โ„ฑ๐‘‡.

Suppose there is a financial market in which ๐‘š+1 securities can be continuously traded. One of them is a bond whose price ๐ต(โ‹…) satisfies d๐ต(๐‘ก)=๐‘Ÿ(๐‘ก)๐ต(๐‘ก)d๐‘ก,(1.1) where ๐‘Ÿ(๐‘ก) is the interest rate of the bond at time ๐‘ก. The other ๐‘š assets are stocks whose dynamicses are subject to the following stochastic differential equations (SDEs): d๐‘†๐‘–(๐‘ก)=๐œ‡๐‘–(๐‘ก)๐‘†๐‘–(๐‘ก)d๐‘ก+๐œŽ๐‘–(๐‘ก)๐‘†๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),(1.2) where ๐œ‡๐‘–(๐‘ก) and ๐œŽ๐‘–(๐‘ก) are called the appreciation rate of return and volatility coefficient of the ๐‘–th stock.

Suppose there is an agent who invests in the bond and stocks, whose decision cannot influence the prices in the financial market. We assume that the trading of the agent is self-financed, that is, there is no infusion or withdrawal of funds over [0,๐‘‡]. We denote by ๐œ‹๐‘–(๐‘ก) the amount that the agent invests in the ๐‘–th stock and by ๐‘ฅ๐œ‹(๐‘ก) the wealth of the agent with an initial endowment ๐‘ฅ0>0. Then the agent has ๐‘ฅ๐œ‹โˆ‘(๐‘ก)โˆ’๐‘š๐‘–=1๐œ‹๐‘–(๐‘ก) savings in a bank. Under the forgoing notations and interpretations, the wealth ๐‘ฅ๐œ‹(โ‹…) is modeled by d๐‘ฅ๐œ‹๎ƒฌ(๐‘ก)=๐‘Ÿ(๐‘ก)๐‘ฅ๐œ‹(๐‘ก)+๐‘š๎“๐‘–=1๎€ท๐œ‡๐‘–๎€ธ๐œ‹(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐‘–๎ƒญ(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๐œŽ๐‘–(๐‘ก)๐œ‹๐‘–(๐‘ก)d๐‘Š๐‘–๐‘ฅ(๐‘ก),๐œ‹(0)=๐‘ฅ0.(1.3)

Generally speaking, it is impossible for the agent to know all the events occurred in the financial market. For instance, if the agent has not enough time or great vigor to observe all the prices of the ๐‘š+1 assets, then the agent will only observe some data of all the prices. Without loss of generality, we denote by ๐’ต๐‘ก the information available to the agent at time ๐‘ก, which is a subfiltration of โ„ฑ๐‘ก. Suppose a process only adapted to ๐’ต๐‘ก is called observable. Therefore, the agent has to choose a portfolio strategy according to the observable filtration ๐’ต๐‘ก. A portfolio strategy ๐œ‹(โ‹…)=(๐œ‹1(โ‹…),โ€ฆ,๐œ‹๐‘š(โ‹…)) is called admissible if ๐œ‹๐‘–(๐‘ก) is a ๐’ต๐‘ก-adapted, square-integrable process with values in ๐‘. The set of the admissible portfolio strategies is denoted by ๐’ฐad.

We give the following hypothesis.(H1)The coefficients ๐‘Ÿ(โ‹…),๐œ‡๐‘–(โ‹…),๐œŽ๐‘–(โ‹…), and ๐œŽ๐‘–(โ‹…)โˆ’1 are uniformly bounded and deterministic functions with values in ๐‘.

For any ๐œ‹(โ‹…)โˆˆ๐’ฐad, (1.3) admits a unique solution under Hypothesis (H1). If we define ๐‘ฃ๐‘–(โ‹…)=๐œŽ๐‘–(โ‹…)๐œ‹๐‘–(โ‹…), then (1.3) is rewritten as d๐‘ฅ๐‘ฃ๎ƒฌ(๐‘ก)=๐‘Ÿ(๐‘ก)๐‘ฅ๐‘ฃ(๐‘ก)+๐‘š๎“๐‘–=1๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๐‘ฃ(๐‘ก)๐‘–๎ƒญ(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๐‘ฃ๐‘–(๐‘ก)d๐‘Š๐‘–๐‘ฅ(๐‘ก),๐‘ฃ(0)=๐‘ฅ0.(1.4)

Let ๐œ‰>0 be a given contingent claim, which is a ๐’ต๐‘‡-measurable, square-integrable random variable. Furthermore, we suppose ๐œ‰ is larger than or equal to ๐‘ฅ0eโˆซ๐‘‡0๐‘Ÿ(๐‘ก)d๐‘ก, where the value ๐‘ฅ0eโˆซ๐‘‡0๐‘Ÿ(๐‘ก)d๐‘ก coincides with the amount that the agent would earn when the initial wealth ๐‘ฅ0 was invested in the bond at the interest rate ๐‘Ÿ(โ‹…) for the entire investment period.

Define a cost functional ๐ฝ๎€ท๐‘ฃ(โ‹…);๐‘ฅ0๎€ธ=12๐”ผ||๐‘ฅ๐‘ฃ||(๐‘‡)โˆ’๐œ‰2.(1.5) Note that the above ๐œ‰ can contain ๐”ผ[๐‘ฅ๐‘ฃ(๐‘‡)โˆฃ๐’ต๐‘‡] as a special case. For a priori given initial wealth ๐‘ฅ0, (1.5) measures the risk that the contingent claim ๐œ‰ cannot be reached. The agentโ€™s objective is ๎€ท๐‘ฃmin๐ฝ(โ‹…);๐‘ฅ0๎€ธsubjectto๐‘ฃ(โ‹…)โˆˆ๐’ฐad,(๐‘ฅ๐‘ฃ(โ‹…);๐‘ฃ(โ‹…))satis๏ฌes(1.3)or(1.4).(PIMV)

The above problem formulates a mean-variance hedging problem with partial information. For simplicity, hereinafter we denote it by the notation โ€œProblem (PIMV)โ€, short for the โ€œpartial information mean-variance hedging problemโ€. In particular, if we let โ„ฑ๐‘ก=๐’ต๐‘ก, 0โ‰ค๐‘กโ‰ค๐‘‡, then Problem (PIMV) reduces to the case with complete information. See, for example, Kohlmann and Zhou [1] for more details.

Because the contingent claim ๐œ‰ in (1.5) is random and the initial endowment ๐‘ฅ0 in (1.3) may be a decision, our Problem (PIMV) is distinguished from the existing literature. See, for example, Pham [2], Xiong and Zhou [3], Hu and ร˜ksendal [4], and so forth. Motivated by Problem (PIMV), we study a general linear-quadratic (LQ) optimal control problem with partial information in Section 2. By a combination of the martingale representation theorem, the technique of โ€œcompleting the squareโ€, and conditional expectation, we derive a corresponding optimal control which is denoted by a related optimal state equation, a Riccati differential equation and a backward stochastic differential equation (BSDE). To demonstrate the applications of our results, we work out some partial information LQ examples and obtain some explicitly observable optimal controls by filtering for BSDEs. Also, we establish some backward and forward-backward stochastic differential filtering equations which are different from the classical ones.

In Section 3, we use the result established in Section 2 to derive an optimal portfolio strategy of Problem (PIMV), which is denoted by the sum of a replicating portfolio strategy for the contingent claim ๐œ‰ and a Mertonโ€™s portfolio strategy. To explicitly illustrate Problem (PIMV), we provide a special but nontrivial example in this section. In terms of filtering theory, we derive the corresponding risk measure. Furthermore, we use some numerical simulations and three figures to illustrate the risk measure and the optimal portfolio strategy.

In Section 4, we compare our results with the existing ones.

Finally, for the convenience of the reader, we state a classical filtering equation for SDEs which is used in Section 3 of this paper.

2. An LQ Optimal Control Problem with Partial Information

In this section, we study a partial information LQ optimal control problem, which is a generalization of Problem (PIMV).

Let us now begin to formulate the LQ problem. Consider a stochastic control system d๐‘ฅ๐‘ฃ๎ƒฌ(๐‘ก)=๐ด(๐‘ก)๐‘ฅ๐‘ฃ(๐‘ก)+๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ฃ๐‘–๎ƒญ(๐‘ก)+๐‘”(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๐‘ฃ๐‘–(๐‘ก)d๐‘Š๐‘–๐‘ฅ(๐‘ก),๐‘ฃ(0)=๐‘ฅ0.(2.1) Here ๐‘ฅ๐‘ฃ(t), ๐‘ฅ0, ๐‘ฃ๐‘–(๐‘ก), ๐‘”(๐‘ก)โˆˆ๐‘๐‘›, ๐ด(๐‘ก) and ๐ต๐‘–(๐‘ก)โˆˆ๐‘๐‘›ร—๐‘›; ๐‘ฃ(โ‹…)=(๐‘ฃ1(โ‹…),โ€ฆ,๐‘ฃ๐‘š(โ‹…)) is a control (process) with values in ๐‘๐‘›ร—๐‘š. We suppose ๐‘ฃ(๐‘ก) is ๐’ต๐‘ก-adapted, where ๐’ต๐‘ก is a given subfiltration of โ„ฑ๐‘ก representing the information available to a policymaker at time ๐‘ก. We say that the control ๐‘ฃ(โ‹…) is admissible and write ๐‘ฃ(โ‹…)โˆˆ๐’ฐad if ๐‘ฃ(โ‹…)โˆˆ๐ฟ2๐’ต(0,๐‘‡;๐‘๐‘›ร—๐‘š), that is, ๐‘ฃ(๐‘ก) is a ๐’ต๐‘ก-adapted process with values in ๐‘๐‘›ร—๐‘š and satisfies ๐”ผ๎€œ๐‘‡0||||๐‘ฃ(๐‘ก)2d๐‘ก<+โˆž.(2.2)

The following basic hypothesis will be in force throughout this section.(H2)๐ด(โ‹…), ๐ต๐‘–(โ‹…) are uniformly bounded and deterministic functions, ๐‘ฅ0 is โ„ฑ0-adapted, and ๐‘”(โ‹…)โˆˆ๐ฟ2โ„ฑ(0,๐‘‡;๐‘๐‘›).

For any ๐‘ฃ(โ‹…)โˆˆ๐’ฐad, control system (2.1) admits a unique solution under Hypothesis (H2). The associated cost functional is ๐ฝ๎€ท๐‘ฃ(โ‹…);๐‘ฅ0๎€ธ=12๐”ผ||๐‘ฅ๐‘ฃ||(๐‘‡)โˆ’๐œ‰2,(2.3) where ๐œ‰ is a given โ„ฑ๐‘‡-measurable, square-integrable random variable. The LQ optimal control problem with partial information is ๎€ท๐‘ฃmin๐ฝ(โ‹…);๐‘ฅ0๎€ธsubjectto๐‘ฃ(โ‹…)โˆˆ๐’ฐad,(๐‘ฅ๐‘ฃ(โ‹…);๐‘ฃ(โ‹…))satis๏ฌes(2.1).(PILQ) An admissible control ๐‘ข(โ‹…) is called optimal if it satisfies ๐ฝ๎€ท๐‘ข(โ‹…);๐‘ฅ0๎€ธ=min๐‘ฃ(โ‹…)โˆˆ๐’ฐad๐ฝ๎€ท๐‘ฃ(โ‹…);๐‘ฅ0๎€ธ.(2.4) The solution ๐‘ฅ(โ‹…) and cost functional (2.3) along with ๐‘ข(โ‹…) are called the optimal state and the value function, respectively.

Problem (PILQ) is related to the recent work by Hu and ร˜ksendal [4], where an LQ control for jump diffusions with partial information is investigated. Due to some characteristic setup, our Problem (PILQ) is not covered by [4]. See, for example, Section 4 in this paper for some detailed comments. Since the nonhomogeneous term in the drift of (2.1) is random and the observable filtration ๐’ต๐‘ก is very general, it is not easy to solve Problem (PILQ). To overcome the resulting difficulty, we shall adopt a combination method of the martingale representation theorem, the technique of โ€œcompleting the squareโ€, and conditional expectation. This method is inspired by Kohlmann and Zhou [1], where an LQ control problem with complete information is studied.

To simplify the cost functional (2.3), we define ๐‘ฆ๐‘ฃ(๐‘ก)=๐‘ฅ๐‘ฃ๎€บ(๐‘ก)โˆ’๐”ผ๐œ‰โˆฃโ„ฑ๐‘ก๎€ป.(2.5) Since ๐”ผ[๐œ‰โˆฃโ„ฑ๐‘ก] is an โ„ฑ๐‘ก-martingale, by the martingale representation theorem (see e.g., Liptser and Shiryasyev [5]), there is a unique ๐‘ง๐‘–(โ‹…)โˆˆ๐ฟ2โ„ฑ(0,๐‘‡;R๐‘›) such that ๐”ผ๎€บ๐œ‰โˆฃโ„ฑ๐‘ก๎€ป=๐”ผ๐œ‰+๐‘š๎“๐‘–=1๎€œ๐‘ก0๐‘ง๐‘–(๐‘ )d๐‘Š๐‘–(๐‘ ).(2.6) Applying Itรดโ€™s formula to (2.1) and (2.5)-(2.6), we have d๐‘ฆ๐‘ฃ๎ƒฌ(๐‘ก)=๐ด(๐‘ก)๐‘ฆ๐‘ฃ(๐‘ก)+๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ฃ๐‘–๎ƒญ(๐‘ก)+โ„Ž(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๎€ท๐‘ฃ๐‘–(๐‘ก)โˆ’๐‘ง๐‘–๎€ธ(๐‘ก)d๐‘Š๐‘–๐‘ฆ(๐‘ก),๐‘ฃ(0)=๐‘ฆ0=๐‘ฅ0โˆ’๐”ผ๐œ‰(2.7) with โ„Ž๎€บ(๐‘ก)=๐‘”(๐‘ก)+๐ด(๐‘ก)๐”ผ๐œ‰โˆฃโ„ฑ๐‘ก๎€ป,(2.8) and cost functional (2.3) reduces to ๐ฝ๎€ท๐‘ฃ(โ‹…);๐‘ฆ0๎€ธ=12๐”ผ||๐‘ฆ๐‘ฃ||(๐‘‡)2.(2.9)

Then Problem (PILQ) is equivalent to minimize (2.9) subject to (2.6)-(2.7) and ๐’ฐad. To solve the resulting problem, we first introduce a Riccati differential equation on ๐‘๐‘›ร—๐‘›ฬ‡๐‘ƒ(๐‘ก)+๐‘ƒ(๐‘ก)๐ด(๐‘ก)+๐ด(๐‘ก)๐œ๐‘ƒ(๐‘ก)โˆ’๐‘š๎“๐‘–=1๐‘ƒ(๐‘ก)๐ต๐‘–(๐‘ก)๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐‘ƒ(๐‘ก)=0,๐‘ƒ(๐‘‡)=๐ผ,๐‘ƒ(๐‘ก)>0,0โ‰ค๐‘กโ‰ค๐‘‡.(2.10) Note that (2.7) contains a nonhomogeneous term โ„Ž(โ‹…). For this, we also introduce a BSDE on ๐‘๐‘›, โˆ’d๐›ผ(๐‘ก)=๎‚ป๎‚ธ๐ด(๐‘ก)๐œโˆ’๐‘šโˆ‘๐‘–=1๐‘ƒ(๐‘ก)๐ต๐‘–(๐‘ก)๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๎‚น๐›ผ(๐‘ก)โˆ’๐‘šโˆ‘๐‘–=1๐‘ƒ(๐‘ก)๐ต๐‘–(๐‘ก)๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๎‚ต๐‘ก)+๐‘ƒ(๐‘ก)โ„Ž(๐‘ก)+๐‘šโˆ‘๐‘–=1๐ต๐‘–(๐‘ก)๐‘ง๐‘–(๐‘ก)๎‚ถ๎‚ผd๐‘กโˆ’๐‘šโˆ‘๐‘–=1๐›ฝ๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐›ผ(๐‘‡)=0.(2.11) Assume that the following hypothesis holds.(H3)For any 0โ‰ค๐‘กโ‰ค๐‘‡, ๐ด(๐‘ก)+๐ด(๐‘ก)๐œโ‰ฅ๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐ต๐‘–(๐‘ก)๐œ.(2.12)

Under Hypotheses (H2) and (H3), according to [1, Theoremโ€‰4.2], it is easy to see that (2.10) admits a unique solution, and then (2.11) admits a unique โ„ฑ๐‘ก-adapted solution (๐›ผ(โ‹…),๐›ฝ1(โ‹…),โ€ฆ,๐›ฝ๐‘š(โ‹…)).

For any admissible pair (๐‘ฃ(โ‹…),๐‘ฆ๐‘ฃ(โ‹…)), using Itรดโ€™s formula to (1/2)๐‘ฆ๐‘ฃ(โ‹…)๐œ๐‘ƒ(โ‹…)๐‘ฆ๐‘ฃ(โ‹…)+๐›ผ(โ‹…)๐œ๐‘ฆ๐‘ฃ(โ‹…), integrating from 0 to ๐‘‡, taking the expectations and trying to complete a square, then we have ๐ฝ๎€ท๐‘ฃ(โ‹…);๐‘ฆ0๎€ธ=๐ฝโ„ฑ๎€ท๐‘ฆ0๎€ธ+12๐”ผ๎€œ๐‘‡0๐‘š๎“๐‘–=1๎€ท๐‘ฃ๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)๐œ๎€ท๐‘ฃ๐‘ƒ(๐‘ก)๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)d๐‘ก=๐ฝโ„ฑ๎€ท๐‘ฆ0๎€ธ+12๐”ผ๎€œ๐‘‡0๐‘š๎“๐‘–=1๐”ผ๐‘ฃ๎€บ๎€ท๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)๐œ๎€ท๐‘ฃ๐‘ƒ(๐‘ก)๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ปd๐‘ก,(2.13) where ๐ฝโ„ฑ๎€ท๐‘ฆ0๎€ธ=12๐‘ฆ๐œ0๐‘ƒ(0)๐‘ฆ0+๐‘ฆ๐œ0+1๐›ผ(0)2๐”ผ๎€œ๐‘‡0๎ƒฌ2๐›ผ(๐‘ก)๐œโ„Ž(๐‘ก)โˆ’2๐‘š๎“๐‘–=1๐›ฝ๐‘–(๐‘ก)๐‘ง๐‘–(๐‘ก)+๐‘š๎“๐‘–=1๐‘ง๐‘–(๐‘ก)๐œ๐‘ƒ(๐‘ก)๐‘ง๐‘–โˆ’(๐‘ก)๐‘š๎“๐‘–=1๎€ท๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐›ผ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๐‘ก)โˆ’๐‘ง๐‘–๎€ธ(๐‘ก)๐œ๎€ท๐‘ƒร—๐‘ƒ(๐‘ก)(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐›ผ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๐‘ก)โˆ’๐‘ง๐‘–๎€ธ๎ƒญ(๐‘ก)d๐‘ก,(2.14)๐ฟ๐‘ฃ๐‘–(๐‘ก)=๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐‘ƒ(๐‘ก)๐‘ฆ๐‘ฃ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐›ผ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๐‘ก)โˆ’๐‘ง๐‘–(๐‘ก).(2.15) Since ๐ฝโ„ฑ(๐‘ฆ0) is independent of ๐‘ฃ๐‘–(โ‹…), the integrand in (2.13) is quadratic with respect to ๐‘ฃ๐‘–(โ‹…) and ๐‘ƒ(โ‹…)>0, then it follows from the property of conditional expectation that the minimum of 12๐”ผ๎€œ๐‘‡0๐‘š๎“๐‘–=1๐”ผ๐‘ฃ๎€บ๎€ท๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)๐œ๎€ท๐‘ฃ๐‘ƒ(๐‘ก)๐‘–(๐‘ก)+๐ฟ๐‘ฃ๐‘–๎€ธ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ปd๐‘ก(2.16) over all ๐’ต๐‘ก-adapted ๐‘ฃ๐‘–(๐‘ก) is attained at ๐‘ข๐‘–๎€บ๐ฟ(๐‘ก)=โˆ’๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป=โˆ’๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๎€ฝ๎€บ๐‘ƒ(๐‘ก)๐”ผ๐‘ฆ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บ+๐”ผ๐›ผ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€พโˆ’๐‘ƒ(๐‘ก)โˆ’1๐”ผ๎€บ๐›ฝ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บ๐‘ง+๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป(2.17) where ๐ฟ๐‘–(๐‘ก)=๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐‘ƒ(๐‘ก)๐‘ฆ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๐›ผ(๐‘ก)+๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๐‘ก)โˆ’๐‘ง๐‘–(๐‘ก)(2.18) and ๐‘ฆ(โ‹…) is the solution of the SDE with ๐‘ข๐‘–(โ‹…) replaced by (2.17) ๎ƒฌd๐‘ฆ(๐‘ก)=๐ด(๐‘ก)๐‘ฆ(๐‘ก)+๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ข๐‘–๎ƒญ(๐‘ก)+โ„Ž(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๎€ท๐‘ข๐‘–(๐‘ก)โˆ’๐‘ง๐‘–๎€ธ(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐‘ฆ(0)=๐‘ฆ0=๐‘ฅ0โˆ’๐”ผ๐œ‰.(2.19)

Now, we are in the position to derive an optimal feedback control in terms of the original optimal state variable ๐‘ฅ(โ‹…). Substituting (2.5) into (2.17), we get ๐‘ข๐‘–๎€บ๐ฟ(๐‘ก)=โˆ’๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป=โˆ’๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๎€ฝ๎€บ๐‘ƒ(๐‘ก)๐”ผ๐‘ฅ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บโˆ’๐‘ƒ(๐‘ก)๐”ผ๐œ‰โˆฃ๐’ต๐‘ก๎€ป๎€บ+๐”ผ๐›ผ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€พโˆ’๐‘ƒ(๐‘ก)โˆ’1๐”ผ๎€บ๐›ฝ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บ๐‘ง+๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป,(2.20) where ๐‘ฅ(โ‹…) satisfies the SDE with ๐‘ข๐‘–(โ‹…) replaced by (2.20) ๎ƒฌd๐‘ฅ(๐‘ก)=๐ด(๐‘ก)๐‘ฅ(๐‘ก)+๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ข๐‘–๎ƒญ(๐‘ก)+๐‘”(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๐‘ข๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐‘ฅ(0)=๐‘ฅ0.(2.21)

Furthermore, we define for any 0โ‰ค๐‘กโ‰ค๐‘‡๎€บ๐‘(๐‘ก)=๐”ผ๐œ‰โˆฃโ„ฑ๐‘ก๎€ปโˆ’๐‘ƒ(๐‘ก)โˆ’1๐‘ž๐›ผ(๐‘ก),๐‘–(๐‘ก)=๐‘ง๐‘–(๐‘ก)โˆ’๐‘ƒ(๐‘ก)โˆ’1๐›ฝ๐‘–(๐‘ก).(2.22) Applying Itรดโ€™s formula to (2.6) and (2.10)-(2.11), we can check that (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) is the unique solution of the BSDE ๎ƒฌd๐‘(๐‘ก)=๐ด(๐‘ก)๐‘(๐‘ก)+๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ž๐‘–๎ƒญ(๐‘ก)+๐‘”(๐‘ก)d๐‘ก+๐‘š๎“๐‘–=1๐‘ž๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐‘(๐‘‡)=๐œ‰.(2.23)

Finally, substituting (2.17) into (2.13), we get the value function ๐ฝ๐’ต๎€ท๐‘ฆ0๎€ธ=๐ฝโ„ฑ๎€ท๐‘ฆ0๎€ธ+12๐”ผ๎€œ๐‘‡0๐‘š๎“๐‘–=1๎€ฝ๐”ผ๎€บ๐ฟ๐‘–(๐‘ก)๐œ๐‘ƒ(๐‘ก)๐ฟ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บ๐ฟโˆ’๐”ผ๐‘–(๐‘ก)๐œโˆฃ๐’ต๐‘ก๎€ป๎€บ๐ฟ๐‘ƒ(๐‘ก)๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€พ๐‘‘๐‘ก,(2.24) where ๐ฝโ„ฑ(๐‘ฆ0) and ๐ฟ๐‘–(โ‹…) are defined by (2.14) and (2.18), respectively.

Theorem 2.1. Let Hypotheses (H2) and (H3) hold. Then the optimal control of Problem (PILQ) is ๐‘ข๐‘–(t)=โˆ’๐‘ƒ(๐‘ก)โˆ’1๐ต๐‘–(๐‘ก)๐œ๎€ฝ๐”ผ๎€บ๐‘ƒ(๐‘ก)๐‘ฅ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บโˆ’๐”ผ๐‘(๐‘ก)โˆฃ๐’ต๐‘ก๎€บ๐‘ž๎€ป๎€พ+๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป,(2.25) where ๐‘ฅ(โ‹…) and (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) are the solutions of (2.21) and (2.23), respectively; the corresponding value function is given by (2.24).

Remark 2.2. Note that the dynamics of BSDE (2.23) is similar to control system (2.1) except for the state constraint, which shows a perfect relationship between stochastic control and BSDE. This interesting phenomenon is first found by [1], to our best knowledge. Also, [1] finds that the solution (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) of (2.23) can be regarded as the optimal state-control pair (๐‘ฅ(โ‹…),๐‘ข(โ‹…)) of an LQ control problem with complete information, in which the initial state ๐‘ฅ0 is an additional decision. That is, ๐‘(โ‹…)=๐‘ฅ(โ‹…) and ๐‘ž๐‘–(โ‹…)=๐‘ข๐‘–(โ‹…) with ๐‘ข(โ‹…)=(๐‘ข1(โ‹…),โ€ฆ,๐‘ข๐‘š(โ‹…)). However, this conclusion is not true in our partial information case. For clarity, we shall illustrate it by the following example.

Example 2.3. Without loss of generality, we let Hypothesis (H2) hold and ๐‘›=1 in Problem (PILQ).

Since ๐‘ƒ(โ‹…) defined by (2.10) is a scalar, it is natural that (2.10) admits a unique solution. Consequently, (2.11) also admits a unique solution. Note that Hypothesis (H3) is not used in this setup. Define ฮ”(โ‹…)=๐‘ฅ(โ‹…)โˆ’๐‘(โ‹…).(2.26) From Itรดโ€™s formula, (2.21)-(2.23) and (2.25), we get ๎ƒฏdฮ”(๐‘ก)=๐ด(๐‘ก)ฮ”(๐‘ก)โˆ’๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)2๐”ผ๎€บฮ”(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป+๐‘š๎“๐‘–=1๐ต๐‘–๎€ท๐”ผ๎€บ๐‘ž(๐‘ก)๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ปโˆ’๐‘ž๐‘–๎€ธ๎ƒฐ+(๐‘ก)d๐‘ก๐‘š๎“๐‘–=1๎€ฝ๐ต๐‘–๎€ท๐”ผ๎€บ๐‘ž(๐‘ก)๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ปโˆ’๐‘ž๐‘–๎€ธ(๐‘ก)โˆ’๐ต๐‘–๎€บ(๐‘ก)๐”ผฮ”(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€พd๐‘Š๐‘–ฮ”(๐‘ก),(0)=๐‘ฅ0โˆ’๐‘(0).(2.27) Hereinafter, we set ๎๎€บ๐‘Œ(๐‘ก)=๐”ผ๐‘Œ(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป,๐‘Œ(โ‹…)=๐‘ฅ(โ‹…),๐‘(โ‹…),๐‘ž1(โ‹…),๐‘ž2(โ‹…),๐‘”(โ‹…),๐‘‹(โ‹…)orฮ”(โ‹…),0โ‰ค๐‘กโ‰ค๐‘‡,(2.28) where the signal ๐‘Œ(๐‘ก) is an โ„ฑ๐‘ก-adapted and square-integrable stochastic progress, while the observation is the component of the ๐‘š-dimensional Brownian motion (๐‘Š(โ‹…)). Without loss of generality, we let the observable filtration ๐’ต๐‘ก be ๐’ต๐‘ก๐‘Š=๐œŽ๎€ฝ๎€ท1(๐‘ ),โ€ฆ,๐‘Š๐‘™๎€ธ๎€พ(๐‘ );0โ‰ค๐‘ โ‰ค๐‘ก,0โ‰ค๐‘กโ‰ค๐‘‡,1โ‰ค๐‘™โ‰ค๐‘šโˆ’1.(2.29) In this setting, we call (2.28) the optimal filtering of the signal ๐‘Œ(๐‘ก) with respect to the observable filtration ๐’ต๐‘ก in the sense of square error. See, for example, [5, 6] for more details.

Note that (๐‘Š1(โ‹…),โ€ฆ,๐‘Š๐‘™(โ‹…)) is independent of (๐‘Š๐‘™+1(โ‹…),โ€ฆ,๐‘Š๐‘š(โ‹…)), ๐‘ฅ0 and ๐‘(0) are deterministic. Taking the conditional expectations on both sides of (2.27), we get the optimal filtering equation of ฮ”(๐‘ก) with respect to ๐’ต๐‘กd๎๎ƒฌฮ”(๐‘ก)=๐ด(๐‘ก)โˆ’๐‘š๎“๐‘–=1๐ต๐‘–(๐‘ก)2๎ƒญ๎ฮ”(๐‘ก)d๐‘กโˆ’๐‘™๎“๐‘–=1๐ต๐‘–๎(๐‘ก)ฮ”(๐‘ก)d๐‘Š๐‘–๎(๐‘ก),ฮ”(0)=๐‘ฅ0โˆ’๐‘(0).(2.30) Note that ๎ฮ”(โ‹…) satisfies a homogeneous linear SDE and hence must be identically zero if ๐‘ฅ0=๐‘(0).

Thereby, if the decision ๐‘ฅ0 takes the value ๐‘(0) in Example 2.3, then the next corollary follows from Theorem 2.1.

Corollary 2.4. The optimal control of Example 2.3 is ๐‘ข๐‘–๎€บ๐‘ž(๐‘ก)=๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป,0โ‰ค๐‘กโ‰ค๐‘‡.(2.31) In particular, if ๐’ต๐‘ก=โ„ฑ๐‘ก, 0โ‰ค๐‘กโ‰ค๐‘‡, then it reduces to the case of [1], that is, ๐‘ข๐‘–(โ‹…)=๐‘ž๐‘–(โ‹…).

From Theorem 2.1 and Corollary 2.4, we notice that the optimal control strongly depends on the conditional expectation of (๐‘(๐‘ก),๐‘ž1(๐‘ก),โ€ฆ,๐‘ž๐‘š(๐‘ก)) with respect to ๐’ต๐‘ก, 0โ‰ค๐‘กโ‰ค๐‘‡, where (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) is the solution of BSDE (2.23). Since ๐’ต๐‘ก is very general, the conditional expectation is, in general, infinite dimensional. Then it is very hard to find an explicitly observable optimal control by some usual methods. However, it is well known that such an optimal control plays an important role in theory and reality. For this, we desire to seek some new technique to further research the problem in the rest of this section. Recently, Wang and Wu [7] investigate the filtering of BSDEs and use a backward separation technique to explicitly solve an LQ optimal control problem with partial information. Please refer to Wang and Wu [8] and Huang et al. [9] for more details about BSDEs with partial information. Inspired by [7, 9], we shall apply the filtering of BSDEs to study the conditional expectation mentioned above. Note that there is no general filtering result for BSDEs in the published literature. In the rest of this section, we shall present two examples of such filtering problems. Combining Theorem 2.1 with a property of conditional expectation, we get some explicitly observable optimal controls. As a byproduct, we establish two new kinds of filtering equations, which are called as backward and forward-backward stochastic differential filtering equations. The result enriches and develops the classical filtering-control theory (see e.g., Liptser and Shiryayev [5], Bensoussan [10], Xiong [6], and so on).

Example 2.5. Let Hypothesis (H2) hold, ๐‘›=1, and ๐‘š=2 in Problem (PILQ). Suppose the observable filtration is denoted by ๐’ต๐‘ก๎€ฝ๐‘Š=๐œŽ1๎€พ(๐‘ );0โ‰ค๐‘ โ‰ค๐‘ก,0โ‰ค๐‘กโ‰ค๐‘‡.(2.32)

From Theorem 2.1, the optimal control is ๐‘ข๐‘–(๐‘ก)=โˆ’๐ต๐‘–๎€ฝ๐”ผ๎€บ๐‘ฅ(๐‘ก)(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บ๐‘โˆ’๐”ผ(๐‘ก)โˆฃ๐’ต๐‘ก๎€บ๐‘ž๎€ป๎€พ+๐”ผ๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป=โˆ’๐ต๐‘–[](๐‘ก)ฬ‚๐‘ฅ(๐‘ก)โˆ’ฬ‚๐‘(๐‘ก)+ฬ‚๐‘ž๐‘–(๐‘ก),(2.33) where (๐‘(โ‹…),๐‘ž1(โ‹…),๐‘ž2(โ‹…)) is the unique solution of ๎ƒฌd๐‘(๐‘ก)=๐ด(๐‘ก)๐‘(๐‘ก)+2๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ž๐‘–๎ƒญ(๐‘ก)+๐‘”(๐‘ก)d๐‘ก+2๎“๐‘–=1๐‘ž๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐‘(๐‘‡)=๐œ‰(2.34) and ๐‘ฅ(โ‹…) satisfies (2.21) with ๐‘š=2.

Similar to Example 2.3, the optimal filtering equation of ๐‘ฅ(โ‹…) is ๎ƒฌdฬ‚๐‘ฅ(๐‘ก)=๐ด(๐‘ก)ฬ‚๐‘ฅ(๐‘ก)+2๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ข๐‘–๎ƒญ(๐‘ก)+ฬ‚๐‘”(๐‘ก)d๐‘ก+๐‘ข1(๐‘ก)d๐‘Š1(๐‘ก),ฬ‚๐‘ฅ(0)=๐‘ฅ0.(2.35) We proceed to calculate the optimal filtering of (๐‘(โ‹…),๐‘ž1(โ‹…),๐‘ž2(โ‹…)). Recalling BSDE (2.34) and noting that the observable filtration is ๐’ต๐‘ก, it follows that ๎ƒฌdฬ‚๐‘(๐‘ก)=๐ด(๐‘ก)ฬ‚๐‘(๐‘ก)+2๎“๐‘–=1๐ต๐‘–(๐‘ก)ฬ‚๐‘ž๐‘–๎ƒญ(๐‘ก)+ฬ‚๐‘”(๐‘ก)d๐‘ก+ฬ‚๐‘ž1(๐‘ก)d๐‘Š1๎€บ(๐‘ก),ฬ‚๐‘(๐‘‡)=๐”ผ๐œ‰โˆฃ๐’ต๐‘‡๎€ป.(2.36) As (2.36) is a (non-Markov) BSDE, we call it a backward stochastic differential filtering equation which is different from the classical filtering equation for SDEs. Since ฬ‚๐‘ž2(โ‹…) is absent from the diffusion term in (2.36), then we are uncertain that if (2.36) admits a unique solution (ฬ‚๐‘(โ‹…),ฬ‚๐‘ž1(โ‹…),ฬ‚๐‘ž2(โ‹…)). But, we are sure that it is true in some special cases. See the following example, in which we establish a forward-backward stochastic differential filtering equation and obtain a unique solution of this equation.

Example 2.6. Let all the assumptions hold and ๐‘”(โ‹…)โ‰ก0 in Example 2.5. For simplicity, we set the random variable ๐œ‰=๐‘‹(๐‘‡), where ๐‘‹(โ‹…) is the solution of d๐‘‹(๐‘ก)=๐พ(๐‘ก)๐‘‹(๐‘ก)d๐‘ก+๐‘€1(๐‘ก)d๐‘Š1(๐‘ก)+๐‘€2(๐‘ก)d๐‘Š2๐‘‹(๐‘ก),(0)=๐‘‹0.(2.37) Assume that ๐พ(โ‹…), ๐‘€1(โ‹…), and ๐‘€2(โ‹…) are bounded and deterministic functions with values in ๐‘; ๐‘‹0 is a constant.

Similar to Example 2.5, the optimal control is ๐‘ข๐‘–(๐‘ก)=โˆ’๐ต๐‘–[](๐‘ก)ฬ‚๐‘ฅ(๐‘ก)โˆ’ฬ‚๐‘(๐‘ก)+ฬ‚๐‘ž๐‘–(๐‘ก),(2.38) where (ฬ‚๐‘ฅ(โ‹…),ฬ‚๐‘(โ‹…),ฬ‚๐‘ž1(โ‹…),ฬ‚๐‘ž2๎(โ‹…),๐‘‹(โ‹…)) is the solution of ๎ƒฌdฬ‚๐‘ฅ(๐‘ก)=๐ด(๐‘ก)ฬ‚๐‘ฅ(๐‘ก)+2๎“๐‘–=1๐ต๐‘–(๐‘ก)๐‘ข๐‘–๎ƒญ(๐‘ก)d๐‘ก+๐‘ข1(๐‘ก)d๐‘Š1(๐‘ก),ฬ‚๐‘ฅ(0)=๐‘ฅ0,๎ƒฌ๐ด(2.39)dฬ‚๐‘(๐‘ก)=(๐‘ก)ฬ‚๐‘(๐‘ก)+2๎“๐‘–=1๐ต๐‘–(๐‘ก)ฬ‚๐‘ž๐‘–๎ƒญ(๐‘ก)d๐‘ก+ฬ‚๐‘ž1(๐‘ก)d๐‘Š1๎(๐‘ก),ฬ‚๐‘(๐‘‡)=๐‘‹(๐‘‡),(2.40)

d๎๎๐‘‹(๐‘ก)=๐พ(๐‘ก)๐‘‹(๐‘ก)d๐‘ก+๐‘€1(๐‘ก)d๐‘Š1๎(๐‘ก),๐‘‹(0)=๐‘‹0.(2.41) It is remarkable that (2.35) together with (2.40)-(2.41) is a forward-backward stochastic differential filtering equation. To our best knowledge, this is also a new kind of filtering equation.

We now desire to give a more explicitly observable representation of ๐‘ข๐‘–(โ‹…). Due to the terminal condition of (2.40), we get by Itรดโ€™s formula and the method of undetermined coefficients, ๎ฬ‚๐‘(โ‹…)=ฮฆ(โ‹…)๐‘‹(โ‹…)+ฮจ(โ‹…),ฬ‚๐‘ž๐‘–(โ‹…)=ฮฆ(โ‹…)๐‘€๐‘–(โ‹…).(2.42) Here ๎๐‘‹(โ‹…) is the solution of (2.41), and ๎€œฮจ(๐‘ก)=โˆ’๐‘‡๐‘ก2๎“๐‘–=1eโˆ’โˆซ๐‘ ๐‘ก๐ด(๐‘Ÿ)d๐‘Ÿฮฆ(๐‘ )๐ต๐‘–(๐‘ )๐‘€๐‘–ฮฆ(๐‘ )d๐‘ ,(๐‘ก)=eโˆซ๐‘‡๐‘ก(๐พ(๐‘ )โˆ’๐ด(๐‘ ))๐‘‘๐‘ .(2.43) Thus, the optimal control is ๐‘ข๐‘–(๐‘ก)=โˆ’๐ต๐‘–๎‚ƒ๎๎‚„(๐‘ก)ฬ‚๐‘ฅ(๐‘ก)โˆ’ฮฆ(๐‘ก)๐‘‹(๐‘ก)โˆ’ฮจ(๐‘ก)+ฮฆ(๐‘ก)๐‘€๐‘–(๐‘ก),(2.44) where ๎๐‘‹(โ‹…) satisfies (2.41) and ฬ‚๐‘ฅ(โ‹…) is the solution of dฬ‚๐‘ฅ(๐‘ก)=๎ƒฏ๎ƒฌ๐ด(๐‘ก)โˆ’2๎“๐‘–=1๐ต๐‘–(๐‘ก)2๎ƒญฬ‚๐‘ฅ(๐‘ก)+2๎“๐‘–=1๐ต๐‘–๎€บ(๐‘ก)ฮฆ(๐‘ก)๐‘€๐‘–(๐‘ก)+๐ต๐‘–๎€ป๎ƒฐ+๎‚†๐ต(๐‘ก)(ฮฆ(๐‘ก)ฬ‚๐‘ฅ(๐‘ก)+ฮจ(๐‘ก))d๐‘ก1๎‚ƒ๎๎‚„(๐‘ก)ฮฆ(๐‘ก)๐‘‹(๐‘ก)โˆ’ฬ‚๐‘ฅ(๐‘ก)+ฮจ(๐‘ก)+ฮฆ(๐‘ก)๐‘€1๎‚‡(๐‘ก)d๐‘Š1(๐‘ก),ฬ‚๐‘ฅ(0)=๐‘ฅ0.(2.45) Since ๎๐‘‹(โ‹…) is the solution of (2.41), it is easy to see that the above equation admits a unique solution ฬ‚๐‘ฅ(โ‹…). Now ๐‘ข๐‘–(๐‘ก), 0โ‰ค๐‘กโ‰ค๐‘‡, defined by (2.44) is an explicitly observable optimal control.

Remark 2.7. BSDE theory plays an important role in many different fields. Then we usually treat some backward stochastic systems with partial information. For instance, to get an explicitly observable optimal control in Theorem 2.1, it is necessary to estimate (๐‘(๐‘ก),๐‘ž1(๐‘ก),โ€ฆ,๐‘ž๐‘š(๐‘ก)) depending on the observable filtration ๐’ต๐‘ก. However, there are short of some effective methods to deal with these estimates. In this situation, although the filtering of BSDEs is very restricted, it can be regarded as an alternative technique (just as we see in Examples 2.2-2.3). By the way, the study of Problem (PILQ) motivates us to establish some general filtering theory of BSDEs in future work. To our best knowledge, this is a new and unexplored research field.

3. Solution to the Problem (PIMV)

We now regard Problem (PIMV) as a special case of Problem (PILQ). Consequently, we can apply the result there to solve the Problem (PIMV). From Theorem 2.1, we get the optimal portfolio strategy ๐œ‹โˆ—๐‘–(๐‘ก)=๐œ‹โˆ—๐‘–1(๐‘ก)+๐œ‹โˆ—๐‘–2(๐‘ก)(3.1) with ๐œ‹โˆ—๐‘–1๐œ‡(๐‘ก)=โˆ’๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–(๐‘ก)2๎€ฝ๐”ผ๎€บ๐‘ฅโˆ—(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป๎€บโˆ’๐”ผ๐‘(๐‘ก)โˆฃ๐’ต๐‘ก,๐œ‹๎€ป๎€พโˆ—๐‘–21(๐‘ก)=๐œŽ๐‘–๐”ผ๎€บ๐‘ž(๐‘ก)๐‘–(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป.(3.2) Here (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) and ๐‘ฅโˆ—(โ‹…) are the solutions of ๎ƒฌd๐‘(๐‘ก)=๐‘Ÿ(๐‘ก)๐‘(๐‘ก)+๐‘š๎“๐‘–=1๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๐‘ž(๐‘ก)๐‘–๎ƒญ(๐‘ก)๐‘‘๐‘ก+๐‘š๎“๐‘–=1๐‘ž๐‘–(๐‘ก)d๐‘Š๐‘–(๐‘ก),๐‘(๐‘‡)=๐œ‰,(3.3)d๐‘ฅโˆ—๎ƒฌ(๐‘ก)=๐‘Ÿ(๐‘ก)๐‘ฅโˆ—(๐‘ก)+๐‘š๎“๐‘–=1๎€ท๐œ‡๐‘–๎€ธ๐œ‹(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)โˆ—๐‘–๎ƒญ(๐‘ก)๐‘‘๐‘ก+๐‘š๎“๐‘–=1๐œŽ๐‘–(๐‘ก)๐œ‹โˆ—๐‘–(๐‘ก)d๐‘Š๐‘–๐‘ฅ(๐‘ก),โˆ—(0)=๐‘ฅ0.(3.4) So we have the following theorem.

Theorem 3.1. If Hypothesis (H1) holds, then the optimal portfolio strategy of Problem (PIMV) is given by (3.1).

We now give a straightforward economic interpretation of (3.1). Introduce an adjoint equation โˆ’d๐œƒ(๐‘ )=๐‘Ÿ(๐‘ )๐œƒ(๐‘ )๐‘‘๐‘ +๐‘š๎“๐‘–=1๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ )๐œŽ๐‘–๐œƒ(๐‘ )๐‘–(๐‘ )d๐‘Š๐‘–(๐‘ ),๐œƒ(๐‘ก)=1,0โ‰ค๐‘กโ‰ค๐‘ โ‰ค๐‘‡.(3.5) Applying Itรดโ€™s formula to ๐‘(โ‹…)๐œƒ(โ‹…), ๐‘(๐‘ก)=eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐”ผ๎‚ƒ๐œ‰eโˆ’โˆซ๐‘‡๐‘กโˆ‘๐‘š๐‘–=1(๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ ))d๐‘Š๐‘–โˆซ(๐‘ )โˆ’(1/2)๐‘‡๐‘กโˆ‘๐‘š๐‘–=1((๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/๐œŽ๐‘–(๐‘ ))2๐‘‘๐‘ โˆฃโ„ฑ๐‘ก๎‚„.(3.6) Note that ๐’ต๐‘ก is a subfiltration of โ„ฑ๐‘ก, 0โ‰ค๐‘กโ‰ค๐‘‡. Then we have ๐”ผ๎€บ๐‘(๐‘ก)โˆฃ๐’ต๐‘ก๎€ป=eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )๐‘‘๐‘ ๐”ผ๎‚ƒ๐œ‰eโˆ’โˆซ๐‘‡๐‘กโˆ‘๐‘š๐‘–=1(๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ ))d๐‘Š๐‘–โˆซ(๐‘ )โˆ’(1/2)๐‘‡๐‘กโˆ‘๐‘š๐‘–=1((๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ )))2d๐‘ โˆฃ๐’ต๐‘ก๎‚„,(3.7) which is the partial information option price for the contingent claim ๐œ‰. According to Corollary 2.4, ๐œ‹โˆ—๐‘–2(โ‹…) is the partial information replicating portfolio strategy for the contingent claim ๐œ‰ when the initial endowment ๐‘ฅ0 is the initial option price ๐‘(0). Then ๐œ‹โˆ—๐‘–1(โ‹…) defined by (3.2) is exactly the partial information Mertonโ€™s portfolio strategy for the terminal utility function ๐‘ˆ(๐‘ฅ)=๐‘ฅ2 (see e.g., Merton [11]). That is, the optimal portfolio strategy (3.1) is the sum of the partial information replicating portfolio strategy for the contingent claim ๐œ‰ and the partial information Mertonโ€™s portfolio strategy. Consequently, if the initial endowment ๐‘ฅ0 is different from the initial option price ๐‘(0) necessary to hedge the contingent claim ๐œ‰, then ๐‘ฅ0โˆ’๐‘(0) should be invested according to Mertonโ€™s portfolio strategy.

In particular, suppose the contingent claim ๐œ‰ is a constant. In this case, it is easy to see that the solution (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) of (3.3) is๐‘(๐‘ก)=eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰,๐‘ž๐‘–(๐‘ก)=0,0โ‰ค๐‘กโ‰ค๐‘‡.(3.8) So we have the following corollary.

Corollary 3.2. Let Hypothesis (H1) hold and ๐œ‰ be a constant. Then the optimal portfolio strategy of Problem (PIMV) is ๐œ‹โˆ—๐‘–๐œ‡(๐‘ก)=โˆ’๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–(๐‘ก)2๎‚†๐”ผ๎€บ๐‘ฅโˆ—(๐‘ก)โˆฃ๐’ต๐‘ก๎€ปโˆ’eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰๎‚‡,0โ‰ค๐‘กโ‰ค๐‘‡.(3.9)

Remark 3.3. The solution (๐‘(โ‹…),๐‘ž1(โ‹…),โ€ฆ,๐‘ž๐‘š(โ‹…)) defined by (3.8) has a straightforward interpretation in financial terms. That is, to achieve a deterministic wealth level ๐œ‰ at the terminal time ๐‘‡, the agent should only invest a risk-free asset (a bond) and cannot invest any risky assets (stocks). Therefore, the optimal portfolio strategy obtained in Corollary 3.2 is only the partial information Mertonโ€™s portfolio strategy.

The left part of this section will focus on a special mean-variance hedging problem with partial information. By virtue of filtering theory, we get an explicitly observable optimal portfolio strategy as well as a risk measure. We also plot three figures and give numerical simulations to illustrate the theoretical result.

Example 3.4. Let ๐‘š=2 and all the conditions in Corollary 3.2 hold. Suppose the observable filtration of an agent is ๐’ต๐‘ก๎€ฝ๐‘†=๐œŽ1๎€พ(๐‘ )โˆถ0โ‰ค๐‘ โ‰ค๐‘ก,0โ‰ค๐‘กโ‰ค๐‘‡.(3.10) It implies that the agents can observe all the past prices of ๐‘†1(โ‹…), but due to some limit factors (e.g., bad behavior of the stock ๐‘†2(โ‹…) or time and energy of the investor) they cannot (do not want to) observe ๐‘†2(โ‹…).

Set ๐’ฎ1(โ‹…)=log๐‘†1(โ‹…), where log๐‘ฅ (๐‘ฅ>0) denotes a logarithm function. It follows from Itรดโ€™s formula that d๐’ฎ1๎‚€๐œ‡(๐‘ก)=11(๐‘ก)โˆ’2๐œŽ1(๐‘ก)2๎‚d๐‘ก+๐œŽ1(๐‘ก)d๐‘Š1(๐‘ก).(3.11) Since ๐œ‡1(โ‹…) and ๐œŽ1(โ‹…) are deterministic functions (see Hypothesis (H1)), the above filtration ๐’ต๐‘ก is equivalently rewritten as ๐’ต๐‘ก๎€ฝ๐’ฎ=๐œŽ1๎€พ๎€ฝ๐‘Š(๐‘ )โˆถ0โ‰ค๐‘ โ‰ค๐‘ก=๐œŽ1๎€พ(๐‘ )โˆถ0โ‰ค๐‘ โ‰ค๐‘ก.(3.12)

Similar to Example 2.5, we get from Corollary 3.2๐œ‹โˆ—๐‘–๐œ‡(๐‘ก)=โˆ’๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–(๐‘ก)2๎‚€ฬ‚๐‘ฅโˆ—(๐‘ก)โˆ’eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰๎‚,(3.13) where ฬ‚๐‘ฅโˆ—(โ‹…) is the solution of dฬ‚๐‘ฅโˆ—(๐‘ก)=๎ƒฏ๎ƒฌ๐‘Ÿ(๐‘ก)โˆ’2๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๎‚ถ(๐‘ก)2๎ƒญฬ‚๐‘ฅโˆ—(๐‘ก)+2๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๎‚ถ(๐‘ก)2eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰๎ƒฐโˆ’๐œ‡๐‘‘๐‘ก1(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ1๎‚€(๐‘ก)ฬ‚๐‘ฅโˆ—(๐‘ก)โˆ’eโˆ’โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰๎‚d๐‘Š1(๐‘ก),ฬ‚๐‘ฅโˆ—(0)=๐‘ฅ0.(3.14) Now ๎๐œ‹โˆ—๐‘–(๐‘ก), 0โ‰ค๐‘กโ‰ค๐‘‡, defined by (3.13), is an observable optimal portfolio strategy.

We now calculate the risk measure (or the value function) of the agentโ€™s goal RM2๎€ทโˆถ=2๐ฝ๐‘ข(โ‹…);๐‘ฅ0๎€ธ=min๐‘ฃ(โ‹…)โˆˆ๐’ฐad๎€ท2๐ฝ๐‘ฃ(โ‹…);๐‘ฅ0๎€ธ.(3.15) From (3.14), we derive ๐”ผฬ‚๐‘ฅโˆ—(๐‘ก)=๐‘ฅ0eโˆซ๐‘ก0โˆ‘[๐‘Ÿ(๐‘ )โˆ’2๐‘–=1((๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ )))2]๐‘‘๐‘ ๎€œ+๐œ‰๐‘ก02๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ )๐œŽ๐‘–๎‚ถ(๐‘ )2eโˆซ๐‘ก๐‘ โˆ‘[๐‘Ÿ(๐œˆ)โˆ’2๐‘–=1((๐œ‡๐‘–(๐œˆ)โˆ’๐‘Ÿ(๐œˆ))/(๐œŽ๐‘–(๐œˆ)))2]dโˆซ๐œˆโˆ’๐‘‡๐‘ ๐‘Ÿ(๐œˆ)d๐œˆd๐‘ ,๐”ผฬ‚๐‘ฅโˆ—(๐‘ก)2=๐‘ฅ20eโˆซ๐‘ก0โˆ‘[2๐‘Ÿ(๐‘ )โˆ’2๐‘–=1((๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ )))2โˆ’((๐œ‡2(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ2(๐‘ )))2]d๐‘ +๎€œ๐‘ก0eโˆซ๐‘ก๐‘ โˆ‘[2๐‘Ÿ(๐œˆ)โˆ’2๐‘–=1((๐œ‡๐‘–(๐œˆ)โˆ’๐‘Ÿ(๐œˆ))/(๐œŽ๐‘–(๐œˆ)))2โˆ’((๐œ‡2(๐œˆ)โˆ’๐‘Ÿ(๐œˆ))/(๐œŽ2(๐œˆ)))2]d๐œˆร—eโˆ’โˆซ๐‘‡๐‘ ๐‘Ÿ(๐œˆ)d๐œˆ๐œ‰๎ƒฌ2๎‚ต๐œ‡2(๐‘ )โˆ’๐‘Ÿ(๐‘ )๐œŽ2๎‚ถ(๐‘ )2๐”ผฬ‚๐‘ฅโˆ—+๎‚ต๐œ‡(๐‘ )1(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ1๎‚ถ(๐‘ก)2eโˆ’โˆซ๐‘‡๐‘ ๐‘Ÿ(๐œˆ)d๐œˆ๐œ‰๎ƒญd๐‘ .(3.16) Combining (3.4), (3.14), Itรดโ€™s formula with Lemma A.1, d๎ƒณ๐‘ฅโˆ—(๐‘ก)2=๎ƒฌ๎ƒณ๐‘ฅ2๐‘Ÿ(๐‘ก)โˆ—(๐‘ก)2+๎‚€eโˆซโˆ’2๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰2โˆ’ฬ‚๐‘ฅโˆ—(๐‘ก)2๎‚2๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๎‚ถ(๐‘ก)2๎ƒญ๐œ‡d๐‘ก+21(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ1๎‚€e(๐‘ก)โˆซโˆ’2๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐œ‰โˆ’ฬ‚๐‘ฅโˆ—(๎‚๐‘ก)ฬ‚๐‘ฅโˆ—(๐‘ก)d๐‘Š1(๎„ž๐‘ฅ๐‘ก),โˆ—(0)2=๐‘ฅ20.(3.17) Solving the above equation, ๐”ผ๎„ž๐‘ฅโˆ—(๐‘‡)2=๐‘ฅ20e2โˆซ๐‘‡0๐‘Ÿ(๐‘ )d๐‘ +๎€œ๐‘‡02๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๎‚ถ(๐‘ก)2๎‚ƒ๐œ‰2โˆ’e2โˆซ๐‘‡๐‘ก๐‘Ÿ(๐‘ )d๐‘ ๐”ผฬ‚๐‘ฅโˆ—(๐‘ก)2๎‚„d๐‘ก.(3.18) Applying a property of conditional expectation and throughout integration by parts, we get RM2๎„ž๐‘ฅ=๐”ผโˆ—(๐‘‡)2โˆ’2๐œ‰๐”ผฬ‚๐‘ฅโˆ—(๐‘‡)+๐œ‰2=๎‚€๐œ‰โˆ’๐‘ฅ0eโˆซ๐‘‡0๐‘Ÿ(๐‘ก)d๐‘ก๎‚2๎‚ต๎€œ1โˆ’๐‘‡0๎‚ถ๐œŒ(๐‘ก)d๐‘ก(3.19) with ๐œŒ(๐‘ก)=2๎“๐‘–=1๎‚ต๐œ‡๐‘–(๐‘ก)โˆ’๐‘Ÿ(๐‘ก)๐œŽ๐‘–๎‚ถ(๐‘ก)2eโˆ’โˆซ๐‘ก0[โˆ‘2๐‘–=1((๐œ‡๐‘–(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ๐‘–(๐‘ )))2+((๐œ‡2(๐‘ )โˆ’๐‘Ÿ(๐‘ ))/(๐œŽ2(๐‘ )))2]d๐‘ .(3.20) So we have the following proposition.

Proposition 3.5. The optimal portfolio strategy and the risk measure are given by (3.13) and (3.19), respectively.

To further illustrate the optimal portfolio strategy (3.13) and the risk measure (3.19), we plot three figures and give some numerical results here. Suppose ๐‘Ÿ=0.06, ๐œ‡1=0.12, ๐œ‡2=0.18, ๐œŽ1=0.12, and ๐œŽ2=0.24. Taking ๐‘‡=1 year, we get from (3.19) ๐œ‰=๐‘ฅ0e0.06+1๎‚™13๎€ท1+2eโˆ’(3/4)๎€ธRM.(3.21)

In Figure 1, we let $0โ‰ค๐‘ฅ0โ‰ค$2 million and $0โ‰คRMโ‰ค$ 1 million. The plane explicitly describes the relationship among ๐œ‰, ๐‘ฅ0, and RM. In detail, the bigger ๐‘ฅ0 and RM, the bigger ๐œ‰.

In Figure 2, we let the investment goal of the agent ๐œ‰=$ 1.2 million. The beeline shows that RM is a deceasing function of the initial endowment ๐‘ฅ0. In particular, when ๐‘ฅ0=$ 1.13 million, we have RM=$ 0. This means that to achieve the investment goal $ 1.2 million at the end of one year, the agent only needs to invest $ 1.13 million in the bond at the interest rate 6%; moreover, there is no risk for the investment strategy.

In Figure 3, we let ๐‘ฅ0=$ 1 million. The beeline implies that RM is an increasing function of the investment goal ๐œ‰. Consider now the agent who has an initial endowment ๐‘ฅ0=$ 1 million and wishes to obtain an expected return rate 20% in one year. Taking ๐‘ฅ0=$ 1 million and ๐œ‰=$ 1.2 million, we get RM=$ 0.1112 million, meaning that the risk of the investment goal is as high as 11.12%.

Furthermore, we calculate the Mertonโ€™s portfolio strategy. Let ๐‘Ÿ=0.06, ๐‘‡=1, ๐‘ฅ0=$ 1 million, and ๐œ‰=$ 1.2 million.(1)Set ๐œ‡1=0.12, ๐œ‡2=0.18, ๐œŽ1=0.12, and ๐œŽ2=0.24. By (3.13), the amount of money the agent should invest in the ๐‘–th stock is ๐œ‹โˆ—๐‘–๐œ‡(๐‘ก)=๐‘–โˆ’0.06๐œŽ2๐‘–๎€ท1.2eโˆ’0.06(1โˆ’๐‘ก)โˆ’ฬ‚๐‘ฅโˆ—๎€ธ(๐‘ก).(3.22)

In particular, at the initial time ๐‘ก=0, ๐œ‹โˆ—1(0)=$ 0.5422 million and ๐œ‹โˆ—2(0)=$ 0.2711 million, which implies that the agent needs to invest $ 0.5422 million and $ 0.2711 million in the stocks ๐‘†1(โ‹…) and ๐‘†2(โ‹…), respectively, and invest in the bond for an amount of 1โˆ’(0.5422+0.2711)=$0.1867million.(3.23)(2)Set ๐œ‡1=0.12, ๐œ‡2=0.18, ๐œŽ1=0.12, and ๐œŽ2=0.17. Similarly, we have ๐œ‹โˆ—1(0)=$ 0.5422 million, ๐œ‹โˆ—2(0)=$ 0.5403 million, and (0.5422+0.5403)โˆ’1=$0.0825million,(3.24)

which implies that the agent needs to borrow $ 0.0825 million from a bank and invest the amount $ 1.0825 million in the two stocks ๐‘†1(โ‹…) and ๐‘†2(โ‹…). This is indeed an aggressive policy.

4. Comparison with Existing Results

The subject of stochastic control with partial information has been discussed by many researchers, such as Bensoussan [10] and Xiong and Zhou [3]. Usually, they made one of the following two assumptions: (i) the filtration ๐’ต is a sigma algebra generated by some observable process; (ii) the control systems are Markovian. From this viewpoint, our work cannot be covered by their results.

Note that our work is related to the recent paper by Hu and ร˜ksendal [4]. In what follows, we shall give some detailed comparisons between them.(1)Comparing with [4], the distinctive characteristics of our Problem (PIMV) are the following four points. First, since ๐œ‰ in our cost functional (1.5) is a random variable, then it partly generalizes that of [4]; meanwhile, our cost functional (1.5) can measure the risk that the contingent claim cannot be reached. Second, we give a possible formulation of partial information in the setting of finance and interpret the economic meaning of the optimal portfolio with partial information. Third, in terms of filtering theory, we explicitly compute the observable optimal portfolio strategy and the risk measure of an agent in Example 3.4. Last but not least, we present some numerical results and figures to illustrate the optimal portfolio strategy and the risk measure. Although the example is special, it is nontrivial and contains some filtering techniques. These results show the practical sense of our paper.(2)Since the initial state ๐‘ฅ0 in control system (2.1) may be a decision and ๐œ‰ in cost functional (2.3) is a random variable, our Problem (PILQ) is different from that of [4]. In particular, if the foregoing ๐‘ฅ0 is also a decision, then the partial information optimal control can be denoted by a conditional expectation of the solution of the corresponding adjoint equation (recall Corollary 2.4). Obviously, this is different from [1, 4].(3)Filtering theory plays an important role in optimal control with partial information. To get an explicitly observable optimal control, it is necessary to compute the conditional expectation of the solution of BSDEs. However, it is not doable in most cases. In this paper, we try to solve some stochastic control problems by using filtering theory. As a byproduct, we establish backward and forward-backward stochastic differential filtering equations which are different from the existing literature about filtering theory. Although the technique of solving stochastic control is restricted and the filtering equations are linear, we can regard them as a contribution to filtering-control theory. By the way, the study about Problem (PILQ) motivates us to establish some general filtering theory of BSDEs in future work.

In [4], Hu and ร˜ksendal obtain some optimal controls and value functions with complete and partial information. Note that they only denote the optimal controls by conditional expectation. We also notice there is not any filtering result in [4]. This is different from ours.


A Classical Filtering Equation for SDEs

We present here a classical filtering result for the readersโ€™ convenience which was employed in this paper. For a detailed discussion of filtering, we refer to the books [5, 6].

Consider the following 1-dimensional state and observation equations: ๐œƒ(๐‘ก)=๐œƒ0+๎€œ๐‘ก0โ„Ž(๐‘ )d๐‘ +๐‘ฅ(๐‘ก),๐œ‰(๐‘ก)=๐œ‰0+๎€œ๐‘ก0๎€œ๐‘Ž(๐‘ ,๐œ”)d๐‘ +๐‘ก0๐‘(๐‘ ,๐œ‰)d๐‘Š(๐‘ ).(A.1) Here (๐‘Š(โ‹…)) is a 1-dimensional standard Brownian motion defined on the complete filtered probability space (ฮฉ,โ„ฑ,(โ„ฑ๐‘ก),๐‘ƒ) equipped with a natural filtration โ„ฑ๐‘ก=๐œŽ{๐‘Š(๐‘ );0โ‰ค๐‘ โ‰ค๐‘ก}, โ„ฑ=โ„ฑ๐‘‡, 0โ‰ค๐‘กโ‰ค๐‘‡; ๐‘ฅ(๐‘ก) is an โ„ฑ๐‘ก-martingale; โ„Ž(๐‘ก) is an โ„ฑ๐‘ก-adapted process with โˆซ๐‘‡0|โ„Ž(๐‘ )|d๐‘ <+โˆž; the functional ๐‘(๐‘ก,๐‘ฆ), ๐‘ฆโˆˆ๐ถ๐‘‡, 0โ‰ค๐‘กโ‰ค๐‘‡, is โ„ฌ๐‘ก-measurable.

We need the following hypothesis.(H)For any ๐‘ฆ,๐‘ฆโˆˆ๐ถ๐‘‡, 0โ‰ค๐‘กโ‰ค๐‘‡, the functional ๐‘(๐‘ก,โ‹…) satisfies ||๎€ท๐‘(๐‘ก,๐‘ฆ)โˆ’๐‘๐‘ก,๐‘ฆ๎€ธ||2โ‰ค๐ฟ1๎€œ๐‘ก0๎€บ๐‘ฆ(๐‘ )โˆ’๎€ป๐‘ฆ(๐‘ )2d๐พ(๐‘ )+๐ฟ2๎€บ๐‘ฆ(๐‘ก)โˆ’๎€ป๐‘ฆ(๐‘ก)2,๐‘(๐‘ก,๐‘ฆ)2โ‰ค๐ฟ1๎€œ๐‘ก0๎€ท1+๐‘ฆ(๐‘ )2๎€ธd๐พ(๐‘ )+๐ฟ2๎€ท1+๐‘ฆ(๐‘ก)2๎€ธ,(A.2)

where ๐ฟ1 and ๐ฟ2 are two nonnegative constants and 0โ‰ค๐พ(โ‹…)โ‰ค1 is a nondecreasing right continuous function. Moreover, sup0โ‰ค๐‘กโ‰ค๐‘‡๐”ผ๐œƒ(๐‘ก)2<+โˆž, โˆซ๐‘‡0๐”ผโ„Ž(๐‘ก)2d๐‘ก<+โˆž, โˆซ๐‘‡0๐”ผ๐‘Ž(๐‘ก,๐œ”)2d๐‘ก<+โˆž, ๐‘(๐‘ก,๐‘ฆ)2โ‰ฅ๐ถ>0.

The following result is due to of [5, Theoremโ€‰8.1].

Lemma A.1. Define ฬ‚๐œƒ(t)=๐”ผ[๐œƒ(t)โˆฃโ„ฑ๐œ‰t]. Here ๐œƒ(t) can take ๐œƒ(t), h(t), D(t), a(t,๐œ”), and ๐œƒ(t)a(t,๐œ”); โ„ฑ๐œ‰t=๐œŽ{๐œ‰(s)โˆถ0โ‰ค๐‘ โ‰คt}, 0โ‰คtโ‰คT. If Hypothesis (H) holds, then the optimal nonlinear filtering equation is ฬ‚ฬ‚๐œƒ๐œƒ(๐‘ก)=0+๎€œ๐‘ก0๎๎€œโ„Ž(๐‘ )d๐‘ +๐‘ก0๎‚†๎๎‚ƒ๎„žฬ‚๎‚„๐ท(๐‘ )+๐œƒ(๐‘ )๐‘Ž(๐‘ ,๐œ”)โˆ’๐œƒ(๐‘ )ฬ‚๐‘Ž(๐‘ ,๐œ”)๐‘(๐‘ ,๐œ‰)โˆ’1๎‚‡d๐‘Š(๐‘ ),(A.3) where ๎€œ๐‘Š(๐‘ก)=๐‘ก0d๐œ‰(๐‘ )โˆ’ฬ‚๐‘Ž(๐‘ ,๐œ”)d๐‘ ๐‘(๐‘ ,๐œ‰)(A.4) is a standard Brownian motion (with respect to โ„ฑ๐œ‰๐‘ก), and ๐ท(๐‘ก) is an โ„ฑ๐‘ก-adapted process with ๐ท(๐‘ก)=dโŸจ๐‘ฅ(๐‘ก),๐‘Š(๐‘ก)โŸฉd๐‘ก.(A.5)


The authors would like to thank the anonymous referee and the Editor for their insightful comments for improving the quality of this work. G. Wang acknowledges the financial support from the National Nature Science Foundation of China (11001156), the Natural Science Foundation of Shandong Province, China (ZR2009AQ017), and the Independent Innovation Foundation of Shandong University, China (2010GN063). Z. Wu acknowledges the financial support from the National Natural Science Foundation of China (10921101), the National Basic Research Program of China (973 Program, no. 2007CB814904), the Natural Science Foundation of Shandong Province, China (JQ200801 and 2008BS01024), and the Science Foundation for Distinguished Young Scholars of Shandong University, China (2009JQ004). The material in this paper was partially presented at the 27th Chinese Control Conference, July 16โ€“18, 2008, Kunming, Yunnan, China.


  1. M. Kohlmann and X. Zhou, โ€œRelationship between backward stochastic differential equations and stochastic controls: a linear-quadratic approach,โ€ SIAM Journal on Control and Optimization, vol. 38, no. 5, pp. 1392โ€“1407, 2000. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  2. H. Pham, โ€œMean-variance hedging for partially observed drift processes,โ€ International Journal of Theoretical and Applied Finance, vol. 4, no. 2, pp. 263โ€“284, 2001. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  3. J. Xiong and X. Zhou, โ€œMean-variance portfolio selection under partial information,โ€ SIAM Journal on Control and Optimization, vol. 46, no. 1, pp. 156โ€“175, 2007. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  4. Y. Hu and B. ร˜ksendal, โ€œPartial information linear quadratic control for jump diffusions,โ€ SIAM Journal on Control and Optimization, vol. 47, no. 4, pp. 1744โ€“1761, 2008. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  5. R. S. Liptser and A. N. Shiryayev, Statistics of random processes, Springer, New York, NY, USA, 1977.
  6. J. Xiong, An introduction to stochastic filtering theory, vol. 18, Oxford University Press, Oxford, UK, 2008.
  7. G. Wang and Z. Wu, โ€œKalman-Bucy filtering equations of forward and backward stochastic systems and applications to recursive optimal control problems,โ€ Journal of Mathematical Analysis and Applications, vol. 342, no. 2, pp. 1280โ€“1296, 2008. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  8. G. Wang and Z. Wu, โ€œThe maximum principles for stochastic recursive optimal control problems under partial information,โ€ IEEE Transactions on Automatic Control, vol. 54, no. 6, pp. 1230โ€“1242, 2009. View at: Publisher Site | Google Scholar
  9. J. Huang, G. Wang, and J. Xiong, โ€œA maximum principle for partial information backward stochastic control problems with applications,โ€ SIAM Journal on Control and Optimization, vol. 48, no. 4, pp. 2106โ€“2117, 2009. View at: Publisher Site | Google Scholar | Zentralblatt MATH
  10. A. Bensoussan, Stochastic control of partially observable systems, Cambridge University Press, Cambridge, UK, 1992. View at: Publisher Site | Zentralblatt MATH
  11. R. Merton, โ€œOptimum consumption and portfolio rules in a continuous-time model,โ€ Journal of Economic Theory, vol. 3, no. 4, pp. 373โ€“413, 1971. View at: Publisher Site | Google Scholar | Zentralblatt MATH

Copyright ยฉ 2011 Guangchen Wang and Zhen Wu. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

More related articles

 PDF Download Citation Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles

We are committed to sharing findings related to COVID-19 as quickly as possible. We will be providing unlimited waivers of publication charges for accepted research articles as well as case reports and case series related to COVID-19. Review articles are excluded from this waiver policy. Sign up here as a reviewer to help fast-track new submissions.