Large scale optimization problems, resulting from analysis of enterprise wide optimization & control objectives, are usually daunted by complexity related to size. Model based and goal based decomposition and co-ordination methods help to simplify the complexity by decomposing the larger, central problem into smaller sub-problems that have a greater possibility of being solved to local optimality. A co-ordination/ co-operation between these sub-systems then helps towards recovering the global optimality of the overall system. Such approaches also offer merits towards permitting additional degrees of freedom for solving the local optimization problems. Work in my research group has focused on analyzing approaches to reduce complexity in optimization problems posed in the areas of scheduling and decision support as well as large scale enterprise wide optimization (such as oil & gas production).

The main tenet of good manufacturing practice (GMP) requires that robust process designs are complemented by continuous improvement based manufacturing strategies. In a typical product development and manufacturing lifecycle beginning from bench scale to full scale, a lot of useful data is generated and archived. The wealth of information present in this data could be exploited towards generating an improved understanding of the cause-effect relationships. This improved understanding could be translated in terms of developing more accurate models that would help in generating realistic and optimal operating policies. Work in my research group has focused on exploiting advances in learning theory to develop improved dynamic models and hence better operating strategies/ recipes for batch and continuous process manufacturing.