Feasible Solution of Non-linear Model Predictive Control for Embedded Implementation.
Research Area - Model Predictive Control, FPGA Implementation, Embedded Control
Sayli Patil
Guide : Dr. D. N. Sonawane
Year : 2017-2019
Mail : [email protected]
My Profile
Research Area - Model Predictive Control, FPGA Implementation, Embedded Control
Sayli Patil
Guide : Dr. D. N. Sonawane
Year : 2017-2019
Mail : [email protected]
My Profile
Abstract -
Many processes need to be operated under tight performance specifications. At the same time more and more constraints, and also environmental and safety considerations need to be satisfied. Often, these demands can only be met when process nonlinearities and constraints are explicitly taken into account in the controller design. Nonlinear Model Predictive Control (NMPC) is the extension of the well-established linear model predictive control to the nonlinear world. NMPC gives more accurate predictions, ability to deal with inequality constraints and use of nonlinear model directly. NMPC requires the repeated on-line solution of a nonlinear optimal control problem. So for the application of NMPC one has to solve a nonlinear programming problem (NLP), which is in general computationally expensive. This is one of the key limiting factors for a successful practical application of NMPC.
This research focuses on exploring various toolboxes (e.g. MAPLE Soft, ACADO, SLSQP, etc.) which generates C/C++ codes for NMPC and to propose the computationally and memory efficient solution for embedded implementation by comparing their performances. Attempt will also targeted to write a customized wrappers on top of generated C/C++ codes of NMPC to improve its robustness and numerical stability for real-time control applications.
Many processes need to be operated under tight performance specifications. At the same time more and more constraints, and also environmental and safety considerations need to be satisfied. Often, these demands can only be met when process nonlinearities and constraints are explicitly taken into account in the controller design. Nonlinear Model Predictive Control (NMPC) is the extension of the well-established linear model predictive control to the nonlinear world. NMPC gives more accurate predictions, ability to deal with inequality constraints and use of nonlinear model directly. NMPC requires the repeated on-line solution of a nonlinear optimal control problem. So for the application of NMPC one has to solve a nonlinear programming problem (NLP), which is in general computationally expensive. This is one of the key limiting factors for a successful practical application of NMPC.
This research focuses on exploring various toolboxes (e.g. MAPLE Soft, ACADO, SLSQP, etc.) which generates C/C++ codes for NMPC and to propose the computationally and memory efficient solution for embedded implementation by comparing their performances. Attempt will also targeted to write a customized wrappers on top of generated C/C++ codes of NMPC to improve its robustness and numerical stability for real-time control applications.
