In this paper we describe some recent developments related to the nume
rical inversion of a Laplace Transform function when the data function
is known on the real axis only. This work is a part of a larger resea
rch project that the authors are developing together with others resea
rchers within the activities of the Research Center of Parallel Comput
ing and Supercomputers. This project is devoted to build a mathematica
l software package for the numerical inversion of a Laplace Transform
function that should be used in real applications and that should allo
w the user to get the more appropriate algorithm according to the spec
ific user problem. If we can state that the complex inversion problem
has been mainly solved, this is not true for the real case. In this la
st case it is much more difficult to develop stable algorithms due to
the strong amplification of the errors during the solution procedure.
Briefly, this is due to an intrinsic ill-posedness of the real inversi
on problem. Therefore, our attention has been devoted to the analysis
of some discretization schemes which we consider the most promising on
es in order to point out some computational issues related to the real
inversion problem and to provide useful guidelines in the development
of efficient and stable algorithms.