Section 3.2 of Bect (2010): where the initial law comes into play #
The hybrid dynamics on E = [0;1]: the state moves to the right at constant speed
v > 0 while it is in E⁰ = [0;1), and jumps instantaneously back to 0 as soon as
it hits the guard G = {1}. Starting from x ∈ [0;1), the k-th jump occurs at time
(k - x) / v and the pre-jump state is always 1, so the jump measure of Eq. (7) is
R = Σ_{k ≥ 1} (law of ((1, (k - X₀)/v))).
Two initial laws are considered.
μ₀ = δ₀: the jumps occur deterministically at the timesk / v, soR = Σ_{k ≥ 1} δ_{(1, k/v)}and no mean jump intensity exists (GSHS.Sawtooth.not_exists_meanJumpIntensity_dirac).μ₀ =the uniform law on[0;1]: then, by Eqs. (9)–(11),R(Γ × (0;t]) = v t δ₁(Γ)(GSHS.Sawtooth.uniform_apply) and the mean jump intensity exists and equalsv δ₁(GSHS.Sawtooth.meanJumpIntensity_uniform), with global mean jump intensityr_t(E) = v.
The state space is modelled inside the ambient real line; all the mass of the process
lives in [0;1].
The jump measure R of Eq. (7) for the constant-speed sawtooth process of
Section 3.2, started with initial law μ₀ on [0;1): the (k+1)-st jump happens at
time ((k+1) - x)/v when the process starts at x, and the pre-jump state is 1.
Equations
- GSHS.Sawtooth.jumpMeasure v μ₀ = MeasureTheory.Measure.sum fun (k : ℕ) => MeasureTheory.Measure.map (fun (x : ℝ) => (1, (↑k + 1 - x) / v)) μ₀
Instances For
The uniform initial law #
Eqs. (9)–(11): the expected number of jumps of the sawtooth process during (0;t],
when started from the uniform law on [0;1], equals v t.
Each summand is the length of {x ∈ (0;1) : (k+1) - v t ≤ x}, namely
min 1 (max 0 (v t - k)), and these clamped values add up to v t.
The value of the jump measure on the rectangles Γ × (0;t], for the uniform
initial law: R(Γ × (0;t]) = v t · δ₁(Γ) (Eq. (11)).
Section 3.2, uniform initial law. The mean jump intensity of the sawtooth
process started from the uniform law on [0;1] exists and is the time-independent
measure v δ₁; in particular the global mean jump intensity is r_t(E) = v.
The Dirac initial law: no mean jump intensity #
Expected number of jumps in (0;t] when the process starts deterministically at
0: it is ⌊v t⌋, a genuinely discontinuous function of t.
Section 3.2, deterministic initial law. When the sawtooth process starts at
x = 0, the jumps occur deterministically at the times k / v, and the process admits
no mean jump intensity: the expected jump count jumps by 1 at t = 1/v, whereas
Definition 6 forces it to be continuous.