Estimating local truncation errors for Runge-Kutta methods (Q688617)

The most popular methods of local error estimation for Runge-Kutta methods involve the embedding of two or more methods within a single step, in doing this the extra cost can be considerable. In this paper the authors recall a method of error estimation by \textit{F. Ceschino} and \textit{J. Kuntzmann} [Problèmes différentiels de conditions initiales (1963; Zbl 0126.125)]. The idea is to combine computed \(y\) and \(f\) values at the end of sufficiently many consecutive steps to obtain an approximation to the local truncation error using a Hermite interpolation formula. The authors generalize this procedure to the variable-stepsize case; the aim is to allow stepsizes to vary freely and to compute the variable- stepsize generalization of the truncation error with the linear combination of solution and derivative values appropriate to the situation computed as the solution proceeds. The note contains the results of some experiments performed with several test problems.