Linear approximation of vector functions (Q2516592)

Let \(f(t) = [f_1(t) , f_2(t) ] \) be a continuous function from the real unit interval, \(0 \leq t \leq 1\) into the 2-dimensional space of real numbers \(R^2\). Equip \(R^2\) with its canonical Euclidian norm. The general question, for the theory of best approximation, would be to characterize, and prove properties of, the best approximation(s) to \(f\) from important classes of functions which serve as approximates. This work addresses the general question in the most basic case of identifying the straight lines that best approximate the ``trajectory'', \(f\). That is, the paper finds necessary and sufficient conditions -- for the most basic, non-trivial case -- that a function \(l(t) = [a_1 + b_1t , a_2 + b_2t]\) be a best approximation to \(f\) from the class of functions linear in each coordinate. The paper also presents methods for constructing such best approximations.