System identification and computer simulation - an integrated approach to the modelling of complex systems

System identification and parameter estimation techniques are important tools for the modelling of complex systems. The paper reviews some practical aspects of system identification and provides an account of an integrated approach to system modelling involving the development of efficient and robust identification methods through the use of computer simulation. The methodology is presented using three apparently contrasting fields of application. These involve the non-invasive estimation of cardio-pulmonary quantities in the human subject, the identification of stability and control derivatives in helicopter flight mechanics models, and the study of complex interactions between neural signals within elements of the neuromuscular system. Similarities highlighted in these different applications include problems of inherent system complexity, difficulties caused by working with short data records and complications introduced by other experimental constraints. Problems of nonlinearity and poor signal-to-noise ratios are also discussed. The role of computer simulation techniques in the development of methods to overcome these problems is outlined in each case and the benefits of an integrated approach are emphasised