Stochastic averaging for non-Lipschitz multi-valued stochastic differential equations driven by G-Brownian motion

Min Han, Yong XuarXiv 2020 · DOI · arXiv

Abstract

In this paper, we prove the validity of an averaging principle for multi-valued stochastic differential equations (MSDEs) driven by G-Brownian motion with non-Lipschitz coefficients. The convergence theorem between the solution of the averaged MSDEs and original one was obtained in the sense of p-th moments and also in capacity. Finally, one example is presented to illustrate our theory.

Introduction

In the field of stochastic analysis, researchers frequently encounter systems characterized by both complex constraints and inherent model uncertainty. Multi-valued stochastic differential equations (MSDEs) are a powerful tool for modeling such systems, particularly those involving reflections or non-smooth boundaries, by utilizing subdifferential operators. Simultaneously, the framework of $G$-Brownian motion ($G$-Bm) has emerged to address "volatility uncertainty"—a scenario where the variance of the stochastic noise is not precisely known but is assumed to lie within a specific range.

While averaging principles are well-established for standard stochastic differential equations, extending these principles to MSDEs driven by $G$-Bm presents significant mathematical challenges. This paper addresses these challenges by establishing an averaging principle under generalized non-Lipschitz conditions, significantly broadening the scope of systems that can be simplified through time-averaging.

Problem Formulation and the Non-Lipschitz Condition

The core difficulty in this research lies in the intersection of two complexities: the multi-valued nature of the equations and the non-additive nature of $G$-expectation. Most existing literature on stochastic averaging assumes that the system's coefficients satisfy a Lipschitz condition, which implies a high degree of regularity.

However, many practical models do not meet this strict requirement. The authors introduce a weaker, more general condition (H1) based on a logarithmic modulus of continuity. This condition allows for coefficients that are less smooth than those required by the standard Lipschitz framework, making the theory applicable to a wider array of real-world phenomena.

Methodology

The proof of the averaging principle relies on several advanced mathematical techniques:

  1. Stochastic Averaging: The authors define a scaling parameter $\epsilon > 0$. As $\epsilon$ approaches zero, the fast-oscillating components of the MSDE are replaced by their long-term averages, resulting in a simplified equation.
  2. Yosida Approximation: To manage the multi-valued operator (which is inherently non-smooth and set-valued), the authors employ the Yosida approximation. This technique approximates the multi-valued operator with a sequence of Lipschitz-continuous, single-valued functions, allowing for standard convergence analysis.
  3. G-Expectation Framework: All derivations are performed within the sublinear expectation space. Unlike classical probability, $G$-expectation handles the ambiguity of the underlying distribution, which is necessary for $G$-Brownian motion.

Key Convergence Results

The paper establishes two primary modes of convergence between the original system ($X^\epsilon$) and the averaged system ($Z^\epsilon$):

Convergence in $p$-th Moments

The authors prove that for a sufficiently small $\epsilon$, the expected difference between the original and averaged solutions vanishes in the $2p$-th moment over a finite time interval $[0, T]$. This result ensures that, on average, the simplified model remains a faithful representation of the complex system.

Convergence in Capacity

Beyond moment convergence, the paper demonstrates uniform convergence in capacity. In the $G$-expectation framework, capacity is the analogue of probability under uncertainty. This result implies that the probability of the two systems deviating significantly is negligible as the scaling parameter decreases, providing a robust guarantee for the approximation.

Illustrative Example

To demonstrate the utility of the expanded theory, the authors present an example using sine-series coefficients. They show that while these coefficients do not satisfy the classical Lipschitz conditions, they do satisfy the generalized non-Lipschitz condition introduced in this work. This confirms that the new framework can handle systems that were previously mathematically intractable.

Conclusion and Future Directions

This research provides a rigorous foundation for simplifying complex MSDEs under volatility uncertainty. By relaxing the Lipschitz requirement, the authors have made the averaging principle applicable to a broader class of stochastic models.

Future research may focus on extending these results to infinite time horizons or exploring more efficient numerical schemes for the resulting averaged equations. Additionally, the application of these theoretical results to specific problems in financial mathematics, where volatility ambiguity is a primary concern, remains a promising area for further study.

Summary updated

Abstract

In this paper, we prove the validity of an averaging principle for multi-valued stochastic differential equations (MSDEs) driven by G-Brownian motion with non-Lipschitz coefficients. The convergence theorem between the solution of the averaged MSDEs and original one was obtained in the sense of p-th moments and also in capicity. Finally, one example is presented to illustrate our theory.
BibTeX
@misc{han2020stochastic,
  author = {Min Han and Yong Xu},
  title = {{Stochastic averaging for non-Lipschitz multi-valued stochastic differential equations driven by G-Brownian motion}},
  year = {2020},
  eprint = {2008.07036},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2008.07036},
}
Figure