Relationship between the relaxed ADP problem and the original infinite-horizon optimal control problem

Determine a relation between the relaxed Lyapunov-function-constrained adaptive dynamic programming formulation and the original infinite-horizon optimal control problem, without imposing strong assumptions on the value function.

Background

The paper modifies the original infinite-horizon optimal control problem by imposing a Lyapunov-function-based stability constraint on the actor optimization problem. Although this relaxation guarantees stability and enables online adaptive dynamic programming, the resulting controller is generally only suboptimal for the original problem. The authors explicitly state that, in general, no relation is known between the relaxed formulation and the original Hamilton–Jacobi–Bellman-based formulation unless strong assumptions are made about the value function. Establishing such a relation would clarify when and how the stability-preserving relaxation can approximate or recover solutions of the original optimal control problem.

References

In general, as pointed out above, no relation between a relaxed and the original formulation of eq:HJB-ctrl is known without strong assumptions on the value function.

— Adaptive dynamic programming using Lyapunov function constraints  (2610.12110 - Göhrt et al., 8 Oct 2026) in Section 1, Introduction