Existence of a solution to the Markov renewal programming problem (Q1108206)

The authors' aim is to show that (in their notation) F: \({\mathbb{R}}^ N\to {\mathbb{R}}^ N\) has a fixed point. They correctly prove that F is continuous, but in the final paragraph go on to claim that F is a contraction mapping since \(\| F(V^ 1)-F(V^ 2)\| \leq k\| V^ 1-V^ 2\|\), where \(0<k<1\). However, in their development k will depend on \(V^ 1\), \(V^ 2\) so that the result appears invalid. (Indeed the whole of this paragraph is unconvincing.) Their presentation mirrors the cited paper by \textit{J. F. Shapiro} [ibid. 49, 710-712 (1975; Zbl 0302.90056)] who dealt with a discounted Markov decision process: here, Brouwer's theorem (rather than the contraction mapping principle) is correctly invoked.