Reason: First publication. The tracked energy bound, discharging hypothesis (EB) of the block cascade under local joint coercivity and without any uniform control-moment hypothesis.
Statement
Adopt simultaneously the settings and notation of the block cascade lemma, of the anchored pre-stopping envelope lemma, of the localized joint coercivity lemma and of the post-exit comparison lemma, all formed for one and the same data: the affine-controlled transition-rate family(β0,β1) on l states with compact convex control set A and its transition-rate familyβ with rate bound B, control bound R and state-Lipschitz constant Λb; the horizon T>0; the population cost data(L,G); the solution with regular event Ω0, empirical state measure Σ, control α and system filtration (Ftsys)t∈[0,T]; the realized control α^; the mean-field trajectory pair(S,A) together with the extensions and the stationary co-stateP, so that (S,A,P) is a stationary mean-field triple; the point x0=S0 of the probability simplex, the realized mean-field flow Φ, the choice S∗=S, the deviation Yt=∣Φt−St∣, the energy E, the reals δ>0 and the clipped-out time O; the fluctuation processess, a and z=(s,a); the noise majorant Q≥0 with E[Q4]≤cQκ0N−2; and the quantities Dt, DG, JN with the bounds CD, CDG and the constant C3=C2+C12/(2r0) of the first-order expansion lemma for the recentred N-agent cost. From the block cascade lemma adopt the block data T0, K, tk, the levels Lk, the matched thresholds cE(k)=Lk2/λc and θout(k)=Lk2/(δ2λo), the constant Λ⋆, the clocks σ(k), the good sets Gk, the leave events Dk, the increments ΔkE and the quantity ΞK; from the localized coercivity lemma adopt the radius ρ∗ of its claim 2 and the constant cJ; and from the post-exit comparison lemma adopt Ctg∨, Cns and εtg. Write 1D for the indicator of a set D and E for the expectation, and let CS=eΛbTlK2T be the constant of claim 4 of the extended good-set stopping-time lemma.
Standing hypotheses. Hypothesis (A) and hypothesis (U) of the quadratic growth lemma; hypothesis (JC) of the localized coercivity lemma, with constant cJ>0 and with r0>0 the constant furnished by claim 5 of that lemma together with (U); hypothesis (LipC) of the pathwise tracking lemma; the optimality [A]∈MS0∗ and hypothesis (TG) of the to-go comparison lemma; and the standing hypothesis on S∗ of the block cascade lemma.
Parameters. Let q0>0 be a real number, put N={Q>q0}, and set
1. (Finiteness and the noise tail.)0≤Z≤4R2TN<∞, the set N is an event with P(N)≤cQκ0q0−4N−2, and
Nk=0∑K−1E[1Gk∩NΔkE]≤4R2TcQκ0q0−4N−1.
2. (Pointwise coercivity on the tracked region.) Let k∈{0,…,K−1}, let ω∈Gk∖N and let t∈[tk,T] with t<σ(k)(ω). Then NDt(ω)≥c⋆∣at(ω)∣2.
3. (Terminal control on the surviving set.)E[1GK∣sT∣2]≤2CS2Z+2cQ1/2κ01/2.
4. (Skeleton inequality.) Let ϵ>0 be a real number and let ρG∗>0 be the radius furnished for it by claim 4 of the localized coercivity lemma, and assume in addition ε1+q0≤ρG∗. Then
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.