next up previous
Next: x_lagrange Up: x_lagrange Previous: x_lagrange

3. A plan

Because we're starting with given data points, let's look for a solution in the form of a time-varying linear combination

$ P(t) = p_0 (t) P_0 + \dots + p_n (t) P_n $.

Here let's use $ p_i $ instead of $ f_i $ to emphasize that the functions we seek are polynomials in $ t $. If each $ p_i $ is a polynomial in $ t $ of degree at most $ n $, then $ P (t) $ will be a polynomial curve of degree at most $ n $, as required.

How can we arrange to have $ P(t_0) = P_0 $? Easily: This happens if $ p_0 (t_0) = 1 $ but $ p_1
(t_0) $ $ = $ $ p_2 (t_0) $ $ = $ ...$ = $ $ p_n (t_0) = 0 $, because then at time $ t_0 $ we get

$ P( t_0 ) = 1 \cdot P_0 + 0 \cdot P_1 + \dots + 0 \cdot P_n =
P_0 $, as desired.

Similarly, we need $ p_1 (t) $ to have values at $ t_0 $, $ t_1,\dots, t_n $ of $ 0, 1, 0,\dots, 0 $ respectively. Then we need $ p_2 (t) $ with a similar property, and so on.

The desired values of $ p_0 (t),\dots, p_n
(t) $, more compactly, are

$ p_i (t_i) = 1 $ for each $ i $, and

$ p_i (t_k) = 0 $ for $ k \neq i $.





Kirby A. Baker 2002-02-13