Sabrina Graber

University of Freiburg, Bosch

Tuesday, March 06, 2018, 11:00 - 12:30

Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany

Numerical optimization is the basis for powerful control techniques such as non-linear model predictive control. Compared to most conventional control methods, it allows to take input and state constraints into account. Direct methods first discretize the optimal control problem and then solve the resulting nonlinear programming. Sequential quadratic programming is a framework of algorithms that solve this finite-dimensional optimization problem. The methods are used in the present thesis to perform a time-optimal setpoint change of the torque in a permanent magnet synchronous machine. For this purpose, different time-optimal or respectively approximately time-optimal problem formulations are set up and different sequential quadratic programming methods are tested and compared against each other. This serves as a basis for controlling the machine with the goal of real-time capability.