Definition (Discounted infinite horizon cost problem)
Given a policy γ∈ΓA , β∈(0,1),with the objective of minimizing Jβ(X,γ)=N→∞limExγ[k=0∑N−1βkc(Xk,Uk)]this is known as the Discounted Optimal Control problem.
Lemma (5.5.1)
Let A be a set and {fn} be a sequence of maps s.t. fn:A→R,∀n∈N. Then n→∞limsupx∈Ainffn(x)≤x∈Ainfn→∞limsupfn(x)
Lemma (5.5.2)
Let Vn(x,u)↑V(x,u) pointwise. Suppose that Vn and V are continuous in u for every x, and u∈U(x)=U is compact. Then, n→∞limu∈U(x)minVn(x,u)=u∈U(x)minV(x,u)
Definition (Discounted Cost Optimality Equation)
We define the discounted cost optimality equation (DCOE) as (T(v))(x):=u∈Umin{c(x,u)+βE[JβN−1(x1)∣x0,u0]}
Lemma (5.5.3)
Define T:v↦T(v) as our DCOE
- If v is a measurable R+−valued function under Measurable Selection Conditions such that v≥T(v) then, v(x)≥Jβ(x)
- Let v≤T(v) and n→∞limβnExγ[v(xn)]=0,∀x∈X,γ∈ΓAThen v(x)≤Jβ(x)
Lemma (5.5.4)
If v(x)=T→∞limJβT(x)is so that v=T(v) where, T(v)(x)=c(x,f(x))+βE[v(x1)∣x0=x,u0=f(x0)] is such that with γ={f,f,…}, n→∞limβnExγ[v(xn)]=0then, γ is optimal.