1: \begin{abstract}

2: 

3: It has been proved \cite{CL08} that discrete least squares polynomial approximation performed on \textbf{(Polynomial) Admissible Meshes}, say \textbf{AM}, enjoys a nice property of convergence. \textbf{Optimal AM}s \cite{KRRE11} are AMs which cardinality grows with optimal rate w.r.t. the degree of approximation.

4: 

5: In Section 2 we show that any compact subset of $\R^d$ that is the closure of a bounded star-like Lipschitz domain $\Omega$ such that $\complement\Omega$ has \emph{positive reach} in the sense of Federer \cite{Fed59}, admits an optimal AM. This extends a result of A. Kro\'o \cite{KRRE11} proved for $\mathscr C^2$ star-shaped domains and is closely related to the recent preprint \cite{KR13}.

6: 

7: In Section $3$ we prove constructively the existence of an optimal AM for  $K:=\overline \Omega\subset \R^d$ where $\Omega$ is a bounded $\mathscr C^{1,1}$ domain. This is done recovering a particular multivariate sharp version of the \emph{Bernstein Inequality} via the \emph{distance function}.

8: \end{abstract}

9: