Data driven modeling and optimal manipulator control

In the field of optimal control linear systems are generally desirable. In the case of model predictive control the benefit of a linear system representation is significant, as it implies convexity of the optimization problem and global optimality for every solution. In this context very interesting is the Koopman operator theory that evolves measurements of a system in a linear function space. Motivated by the enormous potential, the proposed thesis will focus on developing data driven Koopman models for predictive control of robotic manipulators with a serial architecture.