Explicit MPC

Multiparametric (mp) programming pre-computes optimal solutions off-line which are functions of parameters, whose values become apparent online. In explicit MPC, mp-programming is used to compute optimal solutions offline and the online computation is reduced to the simple evaluation of an explicitly defined piecewise linear functions.

 

 

The major research focus in this are are:

  • An enumeration-based approach to multi-parametric programming: 
    Here, an algebraic approach is used to obtain all critical regions of the parametric space. To overcome the combinatorial explosion of the active sets in the optimization problem, inherent in any enumeration approach, an implicit pruning algorithm is proposed.
  • Formulation of point location problem (with Prof. Mani Bhushan):  Point location problem is the most critical step in explicit-MPC execution as its complexity mainly influences the hardware requirements in fast sampling applications. In order to reduce the computational time in e-MPC efficient algorithms are required to solve the point location problem.Our group is attempting to use tools from artificial intelligence literature to design efficient means of solving the point location problem.
  • Naval Gas turbine control using explicit-MPC (wtih Prof. P.S.V. Nataraj): Gas turbines represent a fast dynamic system. Optimal constrained control of such systems are challenging since one requires to solve a constrained optimization problem online. This work involves demonstration of e-MPC on an SR-30 gas turbine in the Systems and Control Laboratory at IIT Bombay.