Evaluation of the derivative of a polynomial in Bernstein form (Q2570696)

Let \(p(t)\) be a polynomial of degree less than or equal to \(n\), defined on \([0,1]\). It is well known that in computational geometry and in computer-aided design the derivatives of \(p(t)\) are usually represented in the so called Bernstein form \[ \frac {d^r(p(t)}{dt^r}= \frac {n!}{n-k)!}\sum _{j=0}^n (\Delta ^r\alpha _j)b_j^n(t)), \quad b_j^n(t)=\binom nj t^j(1-t)^{n-j}, \quad j=0,\dots,n, \] where \(b_j^n(t)\) and \(\Delta ^r\) are the Bernstein polynomial of degree \(n\) and the iterated forward difference operator, respectively. This paper is concerned with the error analysis of the algorithm suggested by the above representation, in the case \(r=1\), for computing the first derivative of \(p(x)\). The corresponding running error is also analyzed. The authors compare their results with those corresponding to a polynomial represented in power basis and using Legendre and Chebyshev forms. Numerical experiments illustrate in detail the behavior of the different representations of a polynomial. Since the conversions between the Bernstein and the power bases and between Bernstein and Chebyshev polynomials were already studied in the literature, for the sake of completeness the conversion between Bernstein and Chebyshev polynomials of the first kind is analyzed in an appendix.