DCT Lexicon · Governing Dynamics
Forward–Backward SDEs and Costates
The stochastic maximum principle couples the state, running forward from today, to a costate, running backward from the horizon. The costate is the marginal value of the state: a price that evolves.
DCT Governing Dynamics
In depth
Formal statement
The Hamiltonian is . Optimality couples a forward SDE to a backward SDE for the adjoint pair .
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Pontryagin, Peng and Pardoux
When the diffusion does not depend on the control, an optimal control maximises the Hamiltonian along the solution of the coupled FBSDE; when it does, Peng's maximum principle adds a second-order adjoint equation. Pardoux–Peng give existence and uniqueness for Lipschitz backward equations; where the value function is smooth, .
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In DCT
The costate is the time-varying price of a unit of each state: what a dollar of liquidity or a point of efficiency is worth at each date. Its path over the programme shows when a unit of each resource matters most.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Limits
Fully coupled FBSDEs may have no solution over long horizons without monotonicity conditions. With a control-dependent diffusion and a nonconvex control set, the first-order adjoint is not enough.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Go next
- HJB Equations and Viscosity Solutions: Where the value is smooth, the costate is its gradient.
- Lagrangian Dual Coordination: Costates are dual prices, dynamic Lagrange multipliers.
- Belief-State Control: Under partial observation the adjoint lives on the belief space.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
References
- Pardoux & Peng (1990)
- Peng (1990), A general stochastic maximum principle
- Yong & Zhou (1999), Stochastic Controls: Hamiltonian Systems and HJB Equations
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In the book: Vol. II, Ch. 6: Optimal Control of Enterprise Systems · Vol. II, Ch. 8: Hamilton–Jacobi–Bellman Enterprise Framework · Vol. II, Ch. 9: Stochastic Enterprise Optimization
Related: HJB Equations and Viscosity Solutions · Lagrangian Dual Coordination · Belief-State Control
GD.13


