A neural approximation to the explicit solution of constrained linear MPC

The solution to constrained linear model predictive control (MPC) problems can be pre-computed off-line in an explicit form as a piecewise affine (PWA) state feedback law defined on polyhedral regions of the state space. Even though real-time optimization is avoided, implementation of the PWA state-feedback law may still require a significant amount of computation due to the problem of determining which polyhedral region the state lies in. In this paper, a neural network approach to this problem is investigated.