Abstract
We estimate the variance of the value function for a random optimal control problem. The value function is the solution w ϵ of a Hamilton–Jacobi equation with random Hamiltonian H(p,x,ω)=K(p)−V(x/ϵ,ω) in dimension d≥2. It is known that homogenization occurs as ϵ→0, but little is known about the statistical fluctuations of w ϵ. Our main result shows that the variance of the solution w ϵ is bounded by O(ϵ/|logϵ|). The proof relies on a modified Poincaré inequality of Talagrand.
Similar content being viewed by others
References
Alexander, K., Zygouras, N.: Subgaussian concentration and rates of convergence in directed polymers. Preprint. arXiv:1204.1819 (2012)
Armstrong, S., Cardaliaguet, P., Souganidis, P.E.: Error estimates and convergence rates for the stochastic homogenization of Hamilton–Jacobi equations. Preprint. arXiv:1206.2601 (2012)
Armstrong, S., Souganidis, P.E.: Stochastic homogenization of level-set convex Hamilton–Jacobi equations. Preprint. arXiv:1203.6303 (2012)
Balázs, M., Quastel, J., Seppäläinen, T.: Fluctuation exponent of the KPZ/stochastic burgers equation. J. Am. Math. Soc. 24, 683–708 (2011)
Bardi, M., Capuzzo-Dolcetta, I.: Optimal Control and Viscosity Solutions of Hamilton–Jacobi–Bellman Equations. Birkhäuser, Basel (1997)
Benaïm, M., Rossignol, R.: Exponential concentration for first passage percolation through modified Poincaré inequalities. Ann. Inst. Henri Poincaré, B Prob. Stat. 44, 544–573 (2008)
Benjamini, I., Kalai, G., Schramm, O.: First passage percolation has sublinear distance variance. Ann. Probab. 31, 1970–1978 (2003)
Corwin, I.: The Kardar–Parisi–Zhang equation and universality class. Random Matrices: Theory Appl. 1, 1130001 (2012)
Johansson, K.: Shape fluctuations and random matrices. Commun. Math. Phys. 209, 437–476 (2000)
Kosygina, E., Varadhan, S.R.S.: Homogenization of Hamilton–Jacobi–Bellman equations with respect to time-space shifts in a stationary ergodic medium. Commun. Pure Appl. Math. 61, 816–847 (2008)
Kosygina, E., Rezakhanlou, F., Varadhan, S.R.S.: Stochastic homogenization of Hamilton–Jacobi–Bellman equations. Commun. Pure Appl. Math. 59, 1489–1521 (2006)
Krug, J., Spohn, H.: Kinetic roughening of growing surfaces. In: Godrèche, C. (ed.) Solids far from Equilibrium. Cambridge University Press, Cambridge (1992)
Lions, P.-L., Souganidis, P.E.: Homogenization of “viscous” Hamilton–Jacobi equations in stationary ergodic media. Commun. Partial Differ. Equ. 30, 335–375 (2005)
Lions, P.-L., Souganidis, P.E.: Stochastic homogenization of Hamilton–Jacobi and “viscous”-Hamilton–Jacobi equations with convex nonlinearities—revisited. Commun. Math. Sci. 8, 637–672 (2010)
Rezakhanlou, F.: Central limit theorem for stochastic Hamilton–Jacobi equations. Commun. Math. Phys. 211, 413–438 (2000)
Rezakhanlou, F., Tarver, J.E.: Homogenization for stochastic Hamilton–Jacobi equations. Arch. Ration. Mech. Anal. 151, 277–309 (2000)
Schwab, R.: Stochastic homogenization of Hamilton–Jacobi equations in stationary ergodic spatio-temporal media. Indiana Univ. Math. J. 58, 527–581 (2009)
Seppäläinen, T.: Scaling for a one-dimensional directed polymer with boundary conditions. Ann. Probab. 40, 19–73 (2012)
Souganidis, P.E.: Stochastic homogenization of Hamilton–Jacobi equations and some applications. Asymptot. Anal. 20, 1–11 (1999)
Talagrand, M.: On Russo’s approximate zero–one law. Ann. Probab. 22, 1576–1587 (1994)
Acknowledgements
The work of J.N. was partially funded by grant DMS-1007572 from the US National Science Foundation.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Proof of Lemma 3.6
The bounds in (3.37) follow from the fact that ℙ is the product measure on \(\varOmega= \{a,b\}^{\mathbb{Z}^{d}_{n}}\), with ℙ(ω(j)=a)=α and ℙ(ω(j)=b)=β. For every nonnegative integrable ψ, (3.37) implies
Similarly \(\mathbb{E}(\psi)\geq C'\mathbb{E}(\psi\circ\phi_{j})\) for all such ψ. □
Proof of Theorem 1.2
Let us define
Then Theorem 1.2 is a slight modification of the following
Theorem A.1
([20], Theorem 1.5)
There is a constant C>0, such that
holds for all f∈L 2(Ω J ).
To derive Theorem 1.2 from this, we start with elementary observation
for C′=min{2α,2β} and C″=max{2α,2β}. Let κ=log(C″/C′)≥0. If \(\log\frac{\|\rho_{j}f\|_{2}}{\|\rho_{j}f\|_{1}} \geq2\kappa\), then
Consequently, Theorem A.1 implies
On the other hand, if \(\log\frac{\|\rho_{j}f\|_{2}}{\|\rho_{j}f\|_{1}} \in[0, 2\kappa)\), then Theorem A.1 implies
Rights and permissions
About this article
Cite this article
Matic, I., Nolen, J. A Sublinear Variance Bound for Solutions of a Random Hamilton–Jacobi Equation. J Stat Phys 149, 342–361 (2012). https://doi.org/10.1007/s10955-012-0590-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-012-0590-y