An integrated batch-to-batch and within batch optimization approach was proposed for real-time optimization of uncertain batch processes. Firstly

the structural information of optimal input trajectory was obtained by optimizing the nominal model

then the whole trajectory was parametrized as scalar variables and sub-trajectories. For the active terminal constraints in the necessary conditions of optimality

the batch-to-batch optimizing approach was taken to satisfy constraint condition. For the sensitivity part

which contained unknown disturbances that entered the optimal feedback

a regression approach was proposed to approximate the optimal input using measurements

hence within batch optimization could be realized. The effectiveness of the proposed integrated optimization approach was verified through the study of a batch reactor.