DCT Lexicon · Governing Dynamics
HJB Equations and Viscosity Solutions
In continuous time, Bellman's recursion becomes a nonlinear partial differential equation for the value function. Viscosity solutions make sense of it when the value function has kinks.
DCT Governing Dynamics
In depth
Formal statement
For and discount rate , the value function solves the Hamilton–Jacobi–Bellman equation:
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Crandall–Lions uniqueness
Under Lipschitz coefficients and a comparison principle, the value function is the unique viscosity solution of the HJB equation, even where it is not differentiable. Monotone, stable, consistent schemes converge to it (Barles–Souganidis).
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In DCT
State constraints (a liquidity floor, a capability floor) create kinks and corners in the value function exactly where the commitments bind. Viscosity solutions keep the theory valid there, and the corners are where the prices of commitments jump.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Limits
Classical solutions rarely exist under constraints or degenerate noise. Numerical schemes must be monotone to converge to the right solution; non-monotone high-order schemes can converge to a wrong one.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Go next
- Forward–Backward SDEs and Costates: The gradient of the value function is the costate of the FBSDE.
- Occupation Measures: The HJB equation's dual is a linear program over occupation measures.
- Dynamic Risk Measures: Risk-sensitive control changes the Hamiltonian to a dynamic risk measure.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
References
- Crandall, Ishii & Lions (1992), User's guide to viscosity solutions
- Fleming & Soner (2006), Controlled Markov Processes and Viscosity Solutions
- Barles & Souganidis (1991)
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In the book: Vol. II, Ch. 8: Hamilton–Jacobi–Bellman Enterprise Framework · Vol. II, Ch. 6: Optimal Control of Enterprise Systems
Related: Forward–Backward SDEs and Costates · Occupation Measures · Dynamic Risk Measures
GD.12


