Optimal control of nonlinear systems with model-reality differences and unmatched terminal constraints

The solution of industrial optimal control problems is heavily dependent upon the model employed for representing the dynamic behaviour and the performance index chosen. For a realistic industrial problem the model and constraints required are often complex and, in addition, the performance index desired may not suit the mathematical techniques available. This has provided the motivation for the development of DISOPE (Dynamic Integrated System Optimisation and Parameter Estimation) [1] which achieves, in an iterative fashion, the correct real optimum solution in spite of differences between the real process and the model used in the computations. Current versions of DISOPE algorithms deal with model-reality differences in model structure, parameters and performance criteria. However, for batch process control, in particular, it would be beneficial if the algorithm could also deal with model-reality differences in terminal conditions and terminal constraints. This paper presents recent preliminary results extending DISOPE to this unmatched terminal constraint situation.