Reason: New lemma for the lower-bound program: the realized control of the controlled N-agent dynamics is progressively measurable with respect to the observation filtration, its truncations are observation-measurable random elements of the control set, and the control energy process is an adapted continuous nondecreasing process. Supplies the measurability layer for the good-set stopping time.
Statement
Adopt the setting, hypotheses and notation of the realized-control lemma: natural numbersN≥1, l≥2, l~≥1, m≥1; a nonempty convex subset A of Euclidean spaceRm which is compact for the topology determined by the Euclidean distance, with a real R≥0 such that ∣a∣≤R for every a∈A, where ∣⋅∣ is the Euclidean norm; a transition-rate familyβ with control set A; an observation-rate familyβ~ with l~ channels; a horizon T>0; an N-agent driving system(Ω,F,P); an A-valuedobservation-driven control policyh=(hk)k≥0 with record spaces Rk(T) (a notation of the policy definition, unrelated to the bound R); a solution(σi,Υυ,α) on [0,T] with regular event Ω0, observation counters N~i,υ indexed by agents i∈{1,…,N} and channels υ∈{1,…,l~}, observation total c~=(c~t)t∈[0,T], observation-event count Kt=c~t, observation filtration (Gt)t∈[0,T] and system filtration (Ftsys)t∈[0,T], both filtrations with time index restricted to [0,T] by clause (vii)(e) of the existence and uniqueness theorem; the event times τj and channels υj (j≥1) of claim 1 of the realized-control lemma; the fixed point a0∈A; the realized control α^:[0,T]×Ω→A of claim 2 there; the set UA of A-valued controls with the metric ρ of the weak metrizability and compactness theorem, formed from the sequence (wr)r∈N fixed in the realized-control lemma; and the convention of claim 3 there by which α^(ω) denotes both the path s↦α^(s,ω) and the element of UA it represents. For t∈[0,T] let B[0,t] be the trace Borel σ-algebra on [0,t] (for t=0 the σ-algebra {∅,{0}}), let B[0,t]⊗Gt be the product σ-algebra on [0,t]×Ω, and write ∫[0,t]⋅ds for the Lebesgue integral over the compact interval[0,t], taken to be 0 for t=0. A real-valued function on Ω is called Gt-measurable when it is measurable with respect to Gt and the Borel σ-algebra of the real line; progressive measurability and stopping times refer to (Gt)t∈[0,T] unless another filtration is named; and for a function θ:Ω→R and a real q we write {θ≤q} for {ω∈Ω:θ(ω)≤q}.
Then the following hold.
1. (Null events.) For every t∈[0,T]: Gt⊆Ftsys; every event of F of probability zero, and in particular Ω∖Ω0, belongs to Gt; and if X is a random variable on (Ω,F,P) and X′ is a Gt-measurable random variable with X(ω)=X′(ω) for every ω outside some event of F of probability zero, then X is Gt-measurable.
2. (Event times and channels.) For every j≥1 and every q∈[0,T] one has {τj≤q}∈Gq. Consequently the function ω↦min(τj(ω),T) is a stopping time of (Gt)t∈[0,T]. Moreover, for every t∈[0,T], every j≥1 and every channel υ∈{1,…,l~},
{τj≤t}∩{υj=υ}∈Gt.
3. (Progressive measurability of the realized control.) Every component α^κ (κ∈{1,…,m}) of the realized control is progressively measurable with respect to (Gt)t∈[0,T], hence also with respect to (Ftsys)t∈[0,T]; in particular, for every t∈[0,T], the restriction of α^κ to [0,t]×Ω is measurable with respect to B[0,t]⊗Gt and the Borel σ-algebra of the real line, and α^κ(t,⋅) is Gt-measurable.
4. (Truncated realized controls.) For t∈[0,T] define α^(t):[0,T]×Ω→A by α^(t)(s,ω)=α^(s,ω) for s≤t and α^(t)(s,ω)=a0 for s>t. Then for every ω∈Ω the path s↦α^(t)(s,ω) is an admissible representative, in the sense of claim 2 of the flow stability lemma, of an element of UA, also written α^(t)(ω), and the paths α^(t)(ω) and α^(ω) agree at every s∈[0,t]; for every ζ∈UA the map ω↦ρ(α^(t)(ω),ζ) is Gt-measurable; and the map ω↦α^(t)(ω) is measurable with respect to Gt and the Borel σ-algebra of the metric space (UA,ρ).
5. (The control energy process.) Let A:[0,T]→A be a map each of whose components is measurable with respect to B[0,T] and the Borel σ-algebra of the real line. Then the family (∣α^(s,⋅)−As∣2)s∈[0,T] is progressively measurable and bounded by 4R2, and
Et(ω)=∫[0,t]α^(s,ω)−As2ds(t∈[0,T],ω∈Ω)
defines a family E=(Et)t∈[0,T] of [0,4R2T]-valued functions on Ω which is progressively measurable with respect to (Gt)t∈[0,T], with EtGt-measurable for every t, and such that 0≤Et(ω)−Et0(ω)≤4R2(t−t0) for all 0≤t0≤t≤T and every ω∈Ω; in particular every path t↦Et(ω) is nondecreasing and continuous on [0,T], the interval and the real line carrying the metric of the real line.
Curated associations between results. These are editable and subjective — they do not replace the dependency graph, which is derived from the references in the text.