Adopt the setting and notation of the controlled N-agent dynamics: natural numbers N≥1, l≥2, l~≥1, m≥1; a nonempty control set A⊆Rm, assumed convex and compact for the topology determined by the Euclidean distance (as required by the realized-control lemma, which is used below); a transition-rate family β with rate bound B; an observation-rate family β~ with l~ channels and rate bound B~; a real number T>0; an N-agent driving system (Ω,F,P); and an A-valued observation-driven control policy h with horizon T. Let a solution on [0,T] for h be given, with regular event Ω0, observation processes Υυ, observation counters N~i,υ, observation total c~, observation-event count Kt=c~t, observation event times τ1<⋯<τKT and channels υ1,…,υKT as in condition 5 of Solution of the Controlled N-Agent Dynamics, observation filtration (Gt)t∈[0,T], and observation record W; the random variables τj and υj (j≥1) are those furnished by claim 1 of the realized-control lemma, so that c~t(ω)≥j if and only if τj(ω)≤t for all ω∈Ω, j≥1 and t∈[0,T], and at every ω∈Ω0 the numbers τj(ω) and υj(ω) with 1≤j≤KT(ω) are the event times and channels of condition 5.
For υ∈{1,…,l~} and u∈[0,T] put c~uυ=∑i=1NN~ui,υ, the channel subtotal, so that c~uυ=NΥuυ by condition 4 of Solution of the Controlled N-Agent Dynamics at every ω∈Ω0, and c~u=∑υ=1l~c~uυ by condition 3 there.
Fix a real number s with 0<s≤T. Let (R(s,l~),R(s,l~)) be the observation record space with horizon s and l~ channels, and let πs:R(T,l~)→R(s,l~) be the prefix map when s<T and the identity map when s=T. Put W(s)=πs∘W. When s<T, let h(s) be the truncated policy and let the restricted solution on [0,s] be the one furnished by claim 2 of that lemma, whose observation record is W(s) and whose observation filtration satisfies Gu(s)=Gu for u∈[0,s]; assume that reconstruction data are fixed for the driving system and the policy h(s) with horizon s (and, when s=T, for the driving system and h with horizon T).
Write N={A∈F:P(A)=0}, σ(W(s))={(W(s))−1(A):A∈R(s,l~)}, 1A for the indicator of a set A, and #E for the number of elements of a finite set E. Measurability of real-valued maps is with respect to the named σ-algebra and the Borel σ-algebra of the real line. Two notational cautions: the letter σ occurs both as a state label and in the state processes σi of Solution of the Controlled N-Agent Dynamics and as the symbol for a generated σ-algebra, the latter always applied to a map and written σ(⋅); and the family N of null events is unrelated to the agent count N. Then the following hold.
1. (The record up to time s is observable.) W(s) is measurable with respect to F and R(s,l~), and σ(W(s))⊆Gs.
2. (Counting identities.) For every ω∈Ω0 and every u∈[0,T],
c~u(ω)=#{j≥1:τj(ω)≤u},c~uυ(ω)=#{j≥1:τj(ω)≤u and υj(ω)=υ}(1≤υ≤l~),
both sets being finite.
3. (The observation processes factor through the record.) For every υ∈{1,…,l~} and every u∈[0,s] there is a map χuυ:R(s,l~)→R, measurable with respect to R(s,l~), such that
Υuυ(ω)=χuυ(W(s)(ω))for every ω∈Ω0.
Explicitly, χuυ(k,t,v)=N1#{j∈{1,…,k}:tj≤u and vj=υ} for a record (k,t,v)∈R(s,l~).
4. (The observation filtration.) Gs is the σ-algebra generated by σ(W(s))∪N. Consequently Gs is W(s)-generated up to null sets in the sense of claim 1 of that lemma, and for every square-integrable random variable X on (Ω,F,P),
E[(X−E[X∣Gs])2]=E[(X−E[X∣σ(W(s))])2].