Best interpolation in a strip (Q1801567)

Let \(0=t_ 1< t_ 2\cdots < t_ n= 1\) be a fixed division of the interval \([0,1]\) and let \(y_ i\), \(i=1,2,\dots,n\) be given real numbers. If \(e\) and \(d\) are linear spline functions at the nodes \(t_ i\) such that \(e_ i=e(t_ i)< y_ i< d(t_ i)= d_ i\), \(i=1,2,\dots,n\), the author considers the problem of finding a function \(f^*\in W^{2,2}\) verifying \(f^*(t_ i)= y_ i\), \(i=1,2,\dots,n\) and \(e(t)\leq f(t)\leq d(t)\), \(t\in [0,1]\) and which minimizes the functional \(\| f''\|\) (\(\|\cdot\|\) -- the \(L^ 2\)-norm). It is proved that there exists a unique cubic spline \(f^*\) verifying these conditions and which depends continuously on the data \((e_ i,y_ i,d_ i)\).