Skip to content

DCT Lexicon · Governing Dynamics

Viability Kernels

The viability kernel is the set of states from which the enterprise can be kept inside its constraints forever, by some admissible policy. Outside it, failure is only a matter of time.

DCT Governing Dynamics

In depth

Formal statement

For dynamics xt+1=F(xt,ut)x_{t+1} = F(x_t,u_t) and a constraint set KK, the viability kernel is the largest subset of KK that admissible controls can keep invariant. It is computed by a decreasing fixed-point iteration.

Viab(K)  =  {x0∈K:∃ (ut) admissible, xt∈K  ∀t≥0}\mathrm{Viab}(K) \;=\; \big\{x_0 \in K : \exists\, (u_t)\ \text{admissible},\ x_t \in K\ \ \forall t \ge 0\big\}K0=K,Kj+1={x∈Kj:∃ u, F(x,u)∈Kj},Viab(K)=⋂jKjK_0 = K,\qquad K_{j+1} = \{x \in K_j : \exists\, u,\ F(x,u) \in K_j\},\qquad \mathrm{Viab}(K) = \bigcap_{j} K_j

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

Viability theorem (Aubin)

For continuous-time differential inclusions x˙∈G(x)\dot x \in G(x) with closed KK and upper semicontinuous, convex, compact-valued GG, KK is viable if and only if G(x)G(x) meets the Bouligand tangent cone at every x∈Kx \in K.

G(x)∩TK(x)≠∅∀x∈KG(x) \cap T_K(x) \neq \varnothing \quad \forall x \in K

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

In DCT

Feasibility today is not enough: a state can satisfy every commitment now and still be outside the kernel, so that every path breaches later. The kernel turns "can we afford this step" into "can we keep affording it", and gives the minimum resources needed to stop in time.

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

Limits

One-step feasibility certifies nothing about the horizon. Kernels are expensive to compute in high dimensions; level-set and grid methods scale exponentially, and robust or stochastic versions shrink the kernel further.

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

Go next

  • Robust Viability: Robust viability keeps the kernel under disturbances (Pathway 5).
  • Bellman Recursion: The kernel is the zero-level set of a Bellman-type value function.
  • Chance Constraints and Certificates: Chance-constrained viability relaxes 'forever' to 'with probability at least 1−α1-\alpha'.

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

References

  • Aubin (1991), Viability Theory
  • Aubin, Bayen & Saint-Pierre (2011), Viability Theory: New Directions

DCT Governing Dynamics: the element's drawer (/governing-dynamics)

In the book: Vol. I, Ch. 14: Mathematical Analysis of the Unified Enterprise Transformation Architecture

Related: Robust Viability · Bellman Recursion · Chance Constraints and Certificates

GD.06