Microcontroller implementation of mixedinteger predictive control

Abstract

The development of efficient methods and also technological advances permit the implementation of predictive control on embedded systems with limited computational power and memory capacity. However, predictive control with a finite number of admissible input values still remains restricted to powerful computational platforms as problem becomes NP-hard. Given the growing computational power of embedded controllers the topic becomes more and more attractive. In the paper the solver based on a standard branch-and-bound method and interior point method is implemented the target system with low power and limited RAM memory. The performance is evaluated in two simple simulation experiments. In the first simulation experiments the goal is to control the level in a MIMO water tank with binary, integer and continuous input signals. The second experiment evaluates the predictive control based on the hybrid model of the two-tank system. © 2016, North Atlantic University Union. All rights Reserved.