Perturbation Analysis (PA) and more particularly Infinitesimal PA (IPA) has been developed as a response to real world problems. Simple IPA calculates a derivative estimate using the holding time sequence under the deterministic similarity assumption that the state part of the trajectory sequence is invariant under perturbation. This is always violated in the limit of an infinitely long sample path or an infinitely large ensemble of finite sample path. The objective of the research is to investigate the following topics: (1) Generalized cut and paste method for IPA analysis, (2) IPA with bias correction and smoothed PA, (3) A-segment methodology and variance reduction, (4) fine tune the software performance of real time computing using PA analysis.
