Ergodic properties of Poisson processes with almost periodic intensity (Q914244): Difference between revisions

Consider a non-homogeneous doubly stochastic Poisson point process (d.s.p.p.p.) \(\Pi\circ A\) on \({\mathbb{R}}\) or \({\mathbb{R}}_+\) driven by an almost periodic (a.p.) intensity function A. Introduce the stationary version \(\Pi \circ A^*\) and the Palm version \(\Pi \circ A^ 0\) corresponding to \(\Pi \circ A^*\). A corresponding stationary random process \(\alpha^*\) and its associated Palm version \(\alpha^ 0\) are constructed for a given deterministic a.p. function \(\alpha\). Let A, \(A^*\) and \(A^ 0\) be the random measures with densities \(\alpha\), \(\alpha^*\) and \(\alpha^ 0\), respectively. Then the random process \(\alpha^*\), the random measure \(A^*\) and the d.s.p.p.p. \(\Pi \circ A^*\) are stationary and ergodic. Let \((T_ i\), \(i\in {\mathbb{N}})\) and \((T^ 0_ i\), \(i\in {\mathbb{N}})\) be the sequence of inter-point distances in the p.p.s. \(\Pi\circ A\) and \(\Pi \circ A^ 0\) restricted to (0,\(\infty)\), respectively. The main results states that for \(j\to \infty\) the distribution of \((T_{i+j}\), \(i\in {\mathbb{N}})\) converges weakly to the distribution of \((T^ 0_ i\), \(i\in {\mathbb{N}})\). Moreover if \(\alpha\) is periodic then we have total variation convergence. The method of proof of the main result involves the invariance property of the Haar measure, Fubini's theorem, and the fact that \((T_ i)\) can be extended to a Markov chain \(((T_ i,K_ i)\), \(i\in {\mathbb{N}})\), where \((K_ i)\) is itself a Markov chain with values in the Bohr compactification \({\hat {\mathbb{R}}}_{\alpha}\) of \({\mathbb{R}}\) relative to \(\alpha\). A limiting distribution of \(K_ i\) as \(i\to \infty\) is derived.