Model predictive control (MPC) is an optimisation-based strategy for high-performance control engineering practice and real-time applications. This method needs to solve online a quadratic programming (QP) problem at each sample time to find optimal control sequence. In this paper, a new optimised MPC architecture is presented, for a gradient-based QP solver to implement linear MPC on a field-programmable gate array platform, which allows obtaining high-quality performances for the real-time control applications. It requires a manual programming of the high-level C/C ++ code in opposition to the other presented approaches, which automatically generates the code. The efficiency of this approach is completed with real time control of the water level of a single tank system running on a Nanoboard 3000XN chip using a conception environment (Altium Designer), while comparing between the MPC and PID controllers.
Read full abstract